Economics and Youth Violence: Crime, Disadvantage, and Community 9780814760239

How do economic conditions such as poverty, unemployment, inflation, and economic growth impact youth violence? Economic

197 107 4MB

English Pages [333] Year 2013

Report DMCA / Copyright


Recommend Papers

Economics and Youth Violence: Crime, Disadvantage, and Community

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Economics and Youth Violence

This page intentionally left blank

Economics and Youth Violence Crime, Disadvantage, and Community

Edited by Richard Rosenfeld, Mark Edberg, Xiangming Fang, and Curtis S. Florence


NEW YORK UNIVERSITY PRESS New York and London © 2013 by New York University All rights reserved References to Internet websites (URLs) were accurate at the time of writing. Neither the author nor New York University Press is responsible for URLs that may have expired or changed since the manuscript was prepared. Library of Congress Cataloging-in-Publication Data Economics and youth violence : crime, disadvantage, and community / edited by Richard Rosenfeld, Mark Edberg, Xiangming Fang, and Curtis S. Florence pages cm Includes bibliographical references and index. ISBN 978-0-8147-8930-8 (hardback) — ISBN 978-0-8147-6059-8 (pb) 1. Youth and violence — Economic aspects. 2. Violence in adolescence — Economic aspects. 3. Juvenile delinquency — Economic aspects. 4. Business cycles. I. Rosenfeld, Richard, editor of compilation. HQ799.2.V56E26 2013 303.60835 — dc23 2013001059 New York University Press books are printed on acid-free paper, and their binding materials are chosen for strength and durability. We strive to use environmentally responsible suppliers and materials to the greatest extent possible in publishing our books. Manufactured in the United States of America c 10 9 8 7 6 5 4 3 2 1 p 10 9 8 7 6 5 4 3 2 1


1 Introduction


Richard Rosenfeld, Curtis S. Florence, Xiangming Fang, and Mark Edberg

Part I: Trends in Macroeconomic Conditions and Youth Violence 2 The Net Effect of the Business Cycle on Crime and Violence


Shawn Bushway, Philip J. Cook, and Matthew Phillips

3 Are the Criminogenic Consequences of Economic Downturns Conditional? Assessing Potential Moderators of the Link between Adverse Economic Conditions and Crime Rates 53 Eric P. Baumer, Richard Rosenfeld, and Kevin T. Wolff

4 Economic Conditions and Violent Victimization Trends among Youth: Guns, Violence, and Homicide, 1973 – 2005


Janet L. Lauritsen, Ekaterina Gorislavsky, and Karen Heimer

Part II: The Neighborhood Context 5 The Nonlinear Effect of Neighborhood Disadvantage on Youth Violence: Neighborhood Effects on Youth Violence


Xiangming Fang, Richard Rosenfeld, Linda L. Dahlberg, and Curtis S. Florence

6 Aggravated Inequality: Neighborhood Economics, Schools, and Juvenile Delinquency


Robert D. Crutchfield and Tim Wadsworth

7 Street Markets, Adolescent Identity, and Violence: A Generative Dynamic


Mark Edberg and Philippe Bourgois >>








own literature review and for shaping the chapters in the current volume. Our first task was to develop an analytic framework to guide our inquiries along the lines of that developed by Brooks-Gunn et al. (1997), but our framework is both narrower in scope and more elaborate in content. In keeping with the objectives of our project, we narrowed the range of macro-social conditions to focus on economic conditions such as poverty, unemployment, economic growth, and wealth inequality. We also narrowed the outcome of concern to violent offending and victimization of children, adolescents, and young adults. But we expanded the mediating domains linking macroeconomic conditions to youth violence to encompass the community (comparable to “neighborhood” in the Brooks-Gunn schema) and situational contexts of youth violence. The situational domain includes family structure and functioning as a major component but also incorporates other relevant micro environments affecting youth violence, such as schools, street gangs, illegal markets, and access to firearms (see table 1.1). Our framework consists of a temporal dimension (long-term versus short-term impacts of economic conditions on youth violence) that is cross-classified with the intervening domains through which the impacts of macroeconomic factors are manifested: societal, community, situational, and individual. The temporal dimension distinguishes economic conditions on the basis of whether they change gradually or rapidly over time. Some macroeconomic factors, such as levels of wealth inequality, change relatively slowly over time, and their impact on youth violence may occur over many years or decades. Such effects are best captured by studies of long-term changes in economic conditions and violence or by cross-sectional studies that reflect the accumulated changes in economic conditions and violence captured at a single point in time. Alternatively, some factors may have an impact on youth violence only when they reach a critical threshold level, and the effective level may differ across communities. Other macroeconomic conditions, such as unemployment, inflation, or consumer confidence, change more rapidly over time, and their effects on violence can be examined in time-series studies covering shorter time periods. The second dimension of the framework specifies the domains through which economic factors may influence youth violence, over both the short and long run. Broad economic phenomena, such as a

Table 1.1. The Impact of Economic Conditions on Youth Violence: A Multilevel Explanatory Framework




national or global recession, may affect entire societies (column one in the figure). Others, such as concentrated poverty or chronic joblessness, affect some communities more than others (column two). Economic conditions also may influence the situational context of youth violence by, for example, affecting firearm ownership and carrying, the salience of street gangs, and markets for illicit drugs (column three). Finally, the economy may affect individual youths in ways that heighten their risk for violence, for example, through its impact on family functioning, community support resources, illicit drug use, or dropping out of school (column four). The short- and long-run impacts of economic conditions are likely to ramify across multiple domains and may differ depending on the race, ethnicity, or immigration status of the affected youth. If nothing else, the analytic framework depicted in table 1.1 should convey the complexity of the manifold connections between economic conditions and youth violence. The panel, in order to make its task manageable, decided to focus its attention on four key subdomains through which the impact of economic conditions on youth violence is manifested: families, schools, community resources, and street markets. These intervening contexts were not intended to exhaust the possible connecting links between economic conditions and youth violence, but they do constitute major pathways through which economic factors are likely to influence levels and patterns of youth violence. The analytic framework proved to be useful for organizing the review and motivating the studies contained in this volume. We recommend continued use of this organizing schema or closely related ones in future reviews intended to summarize prior research and to guide future research on youth violence. The literature review culminated in the following general conclusions: (1) The research literature on macroeconomic factors associated with youth violence is dominated by cross-sectional studies. Further, results from cross-sectional investigations often are not replicated in existing longitudinal research. Perhaps the best example of such divergent results involves the relationship between unemployment and crime. Cross-sectional research reveals a robust, positive relationship between the unemployment rate and crime, whereas the findings of longitudinal studies are mixed, with some studies finding a negative relationship,




The Current Volume Overall, the literature review constitutes a rich resource for the research community by summarizing the major results of, as well as deficits in, recent research on linkages between macroeconomic conditions and youth violence. It also served as a foundation for the current volume, which reviews prior research and introduces new results pertaining to youth violence across multiple domains of influence and over time. We asked the contributors to address important outstanding research issues that emerge from the literature review and to bring new data and insights to bear on these issues from their own research. The volume is organized in three main parts. The three chapters in part 1 focus on the relationship between changing economic conditions and general crime, violence, and youth violence trends over varying time periods. The chapters in part 2 examine patterns of youth violence and its control in relation to community socioeconomic conditions. Part 3 directs attention to how economic conditions influence individual and family factors implicated in youth violence. The final chapter derives policy and prevention implications from the research overviews and the new findings presented in the previous chapters. Part 1 opens with a chapter by Shawn Bushway, Philip Cook, and Matthew Phillips that presents new research results on the relationship between short-run fluctuations in economic conditions and changes in homicide, property crime, and suicide rates over a 76-year period. The results replicate and extend those of a now classic paper by Cook and Zarkin (1985). The authors find that economic downturns result in increases in burglary, robbery, and suicide rates; decreases in motorvehicle theft; and no significant change in homicide rates. An analysis of age-specific arrest rates since 1963 shows basically the same results, except for robbery, which is not significantly related to changes in age-specific arrests. The analysis of arrests indicates that the “procyclical” pattern (crime decreases coincide with economic downturns) for motor-vehicle theft is specific to offenders under the age of 18; no significant results are found for older age groups. The authors view the mechanisms linking cyclical economic change and crime as a “black box” to be filled in by other researchers. The remaining chapters begin to peer inside the black box.




ence report new results from a cross-sectional study of the relationship between neighborhood socioeconomic disadvantage and self-reported violent offending among a nationally representative school-based sample of adolescents. They find that the relationship between disadvantage and adolescent violence is nonlinear: Adolescent violence is relatively low in the least disadvantaged neighborhoods, peaks in more disadvantaged areas, and falls off somewhat in the most disadvantaged areas. The authors recommend that policymakers consider these nonlinear patterns when devising interventions to control youth violence. Robert Crutchfield and Tim Wadsworth present results from their research on youth and adult employment patterns in disadvantaged communities, school performance, and delinquency in a nationally representative sample of adolescents. An important result of this study is that neighborhood socioeconomic disadvantage conditions the effect of school performance on delinquency. Among students who live in disadvantaged areas, better grades in school are associated with higher delinquency rates, a pattern opposite to that found for students from more advantaged areas and suggesting that the relationship between school performance and violence involvement is conditioned by the neighborhood opportunity structure. This finding should concern policymakers and violence-prevention specialists and echoes the conclusion from the literature review that interventions to prevent youth crime and violence should be multivalent, encompassing the community as well as the families and schools of disadvantaged youth. The following chapter by Mark Edberg and Philippe Bourgois considers how the exposure of adolescents to street market contexts (e.g., gangs, drug markets) shapes the connection between violence and identity development — thus increasing the likelihood of violent behavior by youth. Prior research has explored normative aspects of these contexts, particularly a violent “code of the street” (Anderson 1999). The authors, however, focus on the interaction of the street market context as embedded in structural inequality and the adolescent developmental phase of identity formation, concluding that violence comes to represent socially valued personal qualities, which strengthens its connection to positive identity models. The authors also conclude that interventions focusing on the risk factors that contribute to youth involvement in street markets do not sufficiently address the generative dynamic that exists




and short-term economic strains can differentially (1) increase irritability, anger, depression, substance use, harsh parenting, family conflict, and other problems within families, potentially leading to adjustment and academic problems among children; and (2) affect decisions families make about resources they can (and cannot) provide for educational and developmental support. Further, Robert Agnew’s general strain theory (2006) suggests an additional family stressor from a general frustration created by resource deprivation. Matjasko and colleagues propose a model capturing the cyclical dynamic that predicts different family — and thus youth — outcomes by type of economic strain, with implications for intervention. The final chapter by Curtis Florence and Sarah Beth Barnett considers the implications for policy and prevention raised in the preceding chapters. The authors discuss three preconditions for the prevention of youth violence through improvements in economic conditions. First, a strong relationship between youth violence and economic conditions must be established, net of other influences. Second, prevention programs should target the specific economic conditions that affect youth violence, and their effectiveness in improving those conditions must be demonstrated. Third, the economic improvements must be sufficiently large to lead to reductions in youth violence. With respect to the first question, several of the chapters in this volume report statistically significant relationships between economic conditions and crime, including youth violence, but some conditions appear to be more important than others. None of the analyses, for example, shows a direct relationship between unemployment and youth violence. The authors recommend that the impact of youth, adult, and family income levels on youth violence should be given high priority in future research intended to inform prevention policy and programs. The second precondition for policy-relevant research requires that prevention efforts effectively ameliorate the particular economic conditions that influence youth violence. That will mean paying close attention to how economic conditions affect families, schools, and neighborhoods — the more proximate contexts in which youth violence emerges. Finally, the authors maintain that the demonstrated effects of prevention programs must be sufficiently large and sustained to result in appreciable reductions in youth violence. Some of the most effective




2. What are the disaggregated and interactive effects of macroeconomic conditions on youth violence and its community and situational antecedents? 3. How do social institutions influence the level and types of youth violence observed at specific times and places? 4. What is the role of property crime in the generation of and risk for youth violence? 5. To what extent is research on youth violence biased by selection effects? 6. Are data and information on youth violence detailed and timely enough to adequately inform policy responses to youth violence?

Is Macro-Level Research Sufficient? The first question invokes longstanding philosophical and disciplinary debates regarding the adequacy of macro-level theories and research for addressing important questions about human behavior (for an excellent treatment, see Jepperson and Meyer 2011). The debates are framed by two metatheoretical positions: methodological individualism and sociological holism. Methodological individualism, arguably the more dominant orientation in contemporary social science,1 holds that cultural formations, institutions, and social structures are manifested only in the behavior of individuals and that, as such, explanations of human behavior must be grounded in research on individuals. The opposite holistic position, by contrast, posits that supraindividual conditions and processes (e.g., the nuclear family system, market economy, globalization) are sufficient to explain important social outcomes without invoking strong assumptions about the biological characteristics or psychological states of individuals. We have stated these positions in their extreme form, which few social scientists adopt without some qualification. A common response in modern social science to the individualistic-holistic debate is to promote “multilevel” theorizing and research that incorporates macro, micro, and individual properties in the explanation of individual behavior. But the very emphasis on individual outcomes privileges the individualistic position and, by implication, rejects single-level macroscopic explanation as untenable. We fully endorse multilevel theory and research in the study of youth




Elaborating Macro-Level Research If macro-level research on youth violence remains viable and necessary, it also should be expanded and elaborated in at least two ways: greater attention should be directed to explaining the impact of macroeconomic conditions on demographically disaggregated rates of youth violence and to how economic conditions interact in their effects on youth violence. Lauritsen and colleagues provide a useful point of departure for research on disaggregated violence rates in their assessment in this volume of the impact of poverty, unemployment, and consumer sentiment on race- and ethnic-specific rates of violent victimization. Bushway and colleagues find that the impact of the business cycle on motorvehicle theft differs according to the age of the offender (measured by arrests). Baumer and colleagues provide an important example of how some economic conditions (inflation in their study) moderate the influence of other economic factors (unemployment and wages in their study) on youth crime and violence. These results offer a more nuanced view of the relationship between macroeconomic conditions and youth crime and violence than is afforded by research limited to explaining variation in aggregate rates and the additive (versus interactive) effects of multiple macroeconomic conditions. Social Institutions “Macro” social conditions are not limited to the aspects of social structure and culture that are emphasized in most prior research on crime and violence (e.g., labor markets, poverty, inequality, violent conduct norms). Social institutions also matter. Social institutions are the enduring complexes of values, beliefs, organizations, positions, and roles that define a social system and set it apart from others (see Jepperson and Meyer 2011; Messner, Rosenfeld, and Karstedt 2011). The relationship between macroeconomic conditions and rates of youth violence and the effects of the economy on individual youth are likely to vary across differing institutional contexts. Indeed, the very concepts of “youth violence” and “adolescence” as a distinct phase of the developmental process are institutionally determined (see Felson 2002). Neither criminological nor public health research on youth violence has




in this volume). This suggests that property offending may be a significant intervening domain linking the economy to violence. Investigating the causal consequences of property offending and other forms of criminal involvement for youth violence, and the indirect influence of economic conditions on youth violence through property crime, should be high on the future agenda of youth-violence researchers. Selection Effects Perhaps no other issue has bedeviled contemporary research on the community context of youth violence more than that of selection bias (see Duncan, Connell, and Klebanov 1997; Sampson, Morenoff, and Gannon-Rowley 2002). The selection problem is part of a broader class of issues affecting the derivation of causal inferences from nonexperimental data. In the classic experiment, some subjects are randomly allocated to a “treatment” condition and others to a “control” condition. In principle, randomization ensures that the treatment and control subjects are similar on all conditions affecting the outcome, except for the treatment. Randomization is difficult to implement in field research on human behavior, however, and in many instances is simply not possible — especially in the study of risky or violent behavior. It is neither ethical nor practical, for example, to randomly expose children and adolescents to differing levels of poverty, abuse, or neglect to investigate the effects of such community and family conditions on violent behavior. Violence researchers are usually left with few alternatives to the use of observational data in the study of youth violence, that is, observations of violence and its hypothesized antecedents as they naturally occur in real-world environments. The reliance on observational data raises the problem of selection bias.2 Individuals and families are not randomly allocated across communities of different types; they make choices about where to live and, within limits, where to attend school, work, shop, and spend their leisure time. That is, individuals are selected into (and out of) social environments, and those observed in a particular environment may differ in all kinds of ways from those observed elsewhere. If those individual differences are correlated with a specific outcome we want to explain, such as violent behavior, and also with characteristics of the




characteristics and relationship, and features of the event, such as its connection to other crimes and weapon use. Few violence data systems contain this level of detail, and those that do are not nationally representative (e.g., the FBI’s National Incident Based Reporting System) or omit an important type of violence (e.g., the Bureau of Justice Statistics’ National Crime Victimization Survey, which omits lethal violence). Triangulating data across systems is made difficult by the use of differing definitions and reporting jurisdictions. Researchers, policymakers, and the nation would be best served by a single nationally representative violence surveillance system that combines criminal justice and public health data on violent incidents, both lethal and nonlethal, enables comparisons across jurisdictions of varying size, and disseminates the data in small time increments (months or quarters) on a timely basis. Notes 1. For example, Jepperson and Meyer observe, “The idea that structural arguments are at best explanation sketches, to be resolved into and replaced by more adequate individual-level explanations, is now relatively commonplace” (2011, 66). 2. Selection bias and related methodological problems that compromise valid causal inference also occur in randomized field experiments in criminology and other social sciences (Sampson 2010).

References Agnew, Robert. 2006. Pressured into crime: An overview of general strain theory. New York: Oxford University Press. Anderson, Elijah. 1999. Code of the street: Decency, violence, and the moral life of the inner city. New York: Norton. Arvanites, Thomas M., and Robert H. Defina. 2006. Business cycles and street crime. Criminology 44:139 – 64. Brooks-Gunn, Jeanne, Greg J. Duncan, and J. Lawrence Aber, eds. 1997. Neighborhood poverty, vol. 1, Context and consequences for children. New York City: Russell Sage Foundation. Cook, Philip J., and Gary A. Zarkin. 1985. Crime and the business cycle. Journal of Legal Studies 14:115 – 28. Duncan, Greg J., James P. Connell, and Pamela K. Klebanov. 1997. Conceptual and methodological issues in estimating causal effects of neighborhoods and family conditions on individual development. In Jeanne Brooks-Gunn, Greg J. Duncan,


> 23




The chapter begins with a discussion of causal mechanisms linking economic conditions to crime and violence, both overall and for youths. We then investigate the effect of short-term fluctuations in economic activity on crime using the quasi-experimental analysis of the last 13 business cycles (beginning in 1933). Subsequent sections explain our second approach to analyzing short-term fluctuations, a regression analysis on detrended data, and report the results. The regression analysis conveys more statistical power and thereby supports analysis of a variety of outcomes even when available time-series data are of limited duration. The final section concludes.

The Economy’s Influence on Crime and Violence The history of all advanced economies has been characterized by secular economic growth with an overlay of short-term fluctuation, referred to as “cycles” (despite their lack of regularity or predictability). A nonprofit organization, the National Bureau of Economic Research (NBER), assigns quasi-official dates to peaks and troughs in the economic cycle. Each cycle includes two periods, trough to peak (the “expansion” phase) and peak to trough (the “contraction” phase, usually labeled a “recession”). The typical pattern since the 1930s has been of a sustained expansion followed by a relatively short contraction (table 2.1). The NBER’s Business Cycle Dating Committee has this to say: During a recession, a significant decline in economic activity spreads across the economy and can last from a few months to more than a year. Similarly, during an expansion, economic activity rises substantially, spreads across the economy, and usually lasts for several years. The Committee does not have a fixed definition of economic activity. It examines and compares the behavior of various measures of broad activity: real GDP measured on the product and income sides, economywide employment, and real income.2

There has been extensive study of business cycles, in part to determine the degree of synchrony between various indicators of economic activities. Gross domestic product (GDP) and employment are coincident with the cycle, while the stock market averages tend to be “leading




Even this incomplete listing of possible linkages suggests that social scientists should have no trouble providing an explanation for most any sort of empirical finding relating the business cycle to crime. And social scientists have been analyzing the data on this issue for a long time, at least since the mid-19th century (Bonger 1916). Fortunately the empirical findings are quite consistent for some types of crime. Thorsten Sellin’s Research Memorandum on Crime in the Depression (1937) quotes a review of the literature written by Joseph Van Kan in 1903, as follows: Crimes against property find in large measure their indirect causality in bad economic conditions; their direct causality in acute need and even more in chronic misery. . . . Material well-being generally exalts the vital instincts, increases alcohol consumption, and therefore increases crimes against morals. All our literature confirms this fact. . . . As for the question of the extent of the influence of economic factors on offenses against persons, the answers are less uniform. (Quoted in Sellin 1937, 23)

Similarly, Dorothy Swaine Thomas’s classic study Social Aspects of the Business Cycle (1927) reports correlations between detrended crime measures and an indicator of business conditions for Britain, 1857 – 1913, finding strong negative correlations for burglary and robbery and no effect for crimes of violence. These findings have stood up quite well over the ensuing century. One modern analysis focuses on the effects of US business cycles per se. Cook and Zarkin (1985; hereafter “CZ”) utilized annual time series for four types of FBI Index crime for the period 1933 – 1982, using each of the nine cycles during that period as a “trial” in a natural “experiment.” Also reported are simple-regression results on postwar data adjusted for short-term trends (three-year and five-year moving averages). Two measures of economic conditions were used: the unemployment rate and the employment-population ratio.3 Both the nonparametric “natural experiment” study and the regression analyses confirm that burglary and robbery are countercyclical, and homicide rates are not influenced (on balance) by the business cycle. The most surprising result is for auto theft — it is strongly procyclical.4 These analyses are




and the prison population and in one version also control for other business-cycle indicators, namely, the unemployment rate and GDP per capita. The dynamics of their preferred specification is complex, to say the least, since it is estimated in first-difference form with both contemporaneous and lagged ICS, a trend term, and the lagged dependent variable all included. The problem, to put it colloquially, is that in aggregate time-series analysis everything be correlated with everything else, including variables that cannot or are not measured and included. There inevitably is great uncertainty about the proper regression specification when the goal is to isolate the causal effect of a particular variable, such as unemployment or consumer sentiment. It is not only a question of which covariates to include in the specification but also whether to include trend indicators or a lagged dependent variable and how best to account for serial correlation in the error terms. Given the statistical tangle of causal mechanisms that are related to fluctuations in economic activity, generating reliable estimates of their separate contributions from aggregate data appears to us to be beyond reach. That skeptical conclusion takes us back to the simpler task that we set for ourselves, to measure the overall (net) effect of the business cycle on various sorts of crime and violence. The causal ordering is clear, since it is implausible that a change in the crime rate would induce a recession. But in the absence of a structural model, the economic fluctuations that we observe are something of a “black box.” There is no guarantee that the mix of relevant ingredients remains similar from cycle to cycle. As a result, the effect of economic fluctuation on crime may change over time, and the magnitude of the estimated effect therefore depends on the period under consideration. That serves as a caveat with respect to what is otherwise a well-defined goal.

Youth Crime and Violence The effects of the business cycle on crime and violence may be moderated by the age of the perpetrator or the victim. For example, in the case of motor-vehicle theft, we can speculate that adolescents will be less influenced by the market for stolen cars than will adults to the extent that adolescents steal for the purpose of joyriding rather than sale.6




Fig. 2.1

half years since 1969. This increase reflects the longest expansion in the postwar period, a ten-year expansion from March 1991 to March 2001. The up-and-down nature of the business cycle can be illustrated through an examination of two economic indicators, the employment ratio (the ratio of the population employed to the total population) and the unemployment rate (the percentage of people in the labor market who are not working), during this period. Figure 2.1 provides a description of the unemployment rate and the employment ratio from 1948 to 2008. The bars mark the peak years of the business-cycle expansions during this time period. (Since we are working with annual data, the peak year with respect to GDP/capita is not necessarily the same as the year of the peak month.) These two statistical trends are telling similar stories — unemployment increases when employment decreases — in close connection with the trough-peak-trough pattern of the business cycle. Followers of crime trends will know that crime in the United States is not cyclical in this same way but rather tends to exhibit long swings.




Table 2.1. Vital Statistics Homicide and Suicide Movements over 12 Business Cycles Homicide Reference cycle dates Trough


Trough-to-peak interval used in calculations

3/33 6/38 10/45 10/49 5/54 4/58 2/61 11/70 3/75 7/80 11/82 3/91 11/01

5/37 2/45 11/48 7/53 8/57 4/60 12/69 11/73 1/80 7/81 7/90 3/01 12/07

1933–1937 1938–1945 1946–1948 1949–1953 1954–1957 1958–1960 1961–1969 1971–1973 1975–1979 1980–1981 1983–1990 1991–2000 2001–2007

Number of pluses

Annual growth trough to peak (%) –5.9 –2.7 –4.1 –2.9 –2.1 2.2 6.4 3.8 0.0 –3.7 2.2 –6.0

Next year (%) –10.5 12.5 –6.9 0.0 0.0 0.0 7.8 4.1 7.0 –6.8 5.0 18.3

8 ( p-value = .39)


+ + + + + + + + +

Annual growth trough to peak (%) –1.4 –4.4 –1.3 –3.0 –1.0 0.5 0.8 1.3 –1.2 0.8 0.4 –1.8

Next year (%) 2.0 2.7 1.8 0.0 9.2 –1.9 4.5 0.8 –1.7 1.7 –1.6 2.9

+ + + + + + +

+ +

8 ( p-value = .39)

Source: Vital Statistics 1933–2006 Notes: Number of pluses indicates the number of cycles in which growth in the crime rate from trough to peak was less than the rate of growth in the following year. Vital Statistics data are not available for 2007–2008.

rate of growth to be higher during the recession than during the expansion period. Alternatively, if crime is procyclical, we would expect the rate of growth to be higher during the expansion period than during the subsequent recession. Figure 2.3 provides a graphical description of this comparison. For each business cycle (and for each crime), we subtracted the average annual rate of change for the boom period from the average annual rate of change for the bust period. We expect this difference to be negative on average if recessions cause crime to increase and positive if crime increases during boom periods. Murder is acyclical, with roughly equal numbers of business cycles in which the rate of change was higher during the recession and higher during the expansion. In contrast, the rates of growth were higher during the recession for most of the cycles for both robbery and burglary. Finally, auto theft appears to be either acyclical or slightly procyclical,

Table 2.2. UCR Crime Movements over 13 Business Cycles Murder



Auto theft

Troughto-peak inteval used in calculations

Annual growth trough Next to peak year (%) (%)

Annual growth trough Next to peak year + (%) (%)

Annual growth trough Next to peak year + (%) (%)

Annual growth trough Next to peak year + (%) (%)

1933–1937 1938–1945 1946–1948 1949–1953 1954–1957 1958–1960 1961–1969 1971–1973 1975–1979 1980–1981 1983–1990 1991–2000 2001–2007

–2.4 –1.6 –6.7 –0.9 –1.4 3.1 5.4 4.5 0.3 –3.9 1.8 –6.2 0.0

+ + + + +

+ + + + +

Number of pluses

–5.7 16.9 –10.0 –5.8 0.0 –6.0 8.2 4.3 5.2 –7.1 4.3 1.8 –4.4

6 ( p-value = 1.0)


+ + + + +

–13.2 –1.6 –6.6 0.0 –4.9 4.5 12.4 –1.3 –0.3 3.0 2.4 –6.8 –0.1

–1.6 9.8 5.8 4.7 10.9 –3.2 16.0 14.3 15.0 –7.7 6.4 2.4 –1.1

10 ( p-value = .09)

+ + + + +

–6.9 0.7 –1.3 2.4 2.4 6.9 8.3 2.5 –0.3 –2.2 –1.2 –5.8 –0.4

0.8 5.6 6.6 6.4 10.7 2.0 10.2 17.6 11.4 –9.7 1.6 1.8 1.3

11 ( p-value = .02)

+ + + + + +

–10.0 3.2 –15.7 6.8 8.2 4.1 11.4 –1.9 1.6 –5.6 6.2 –5.1 –2.8

–13.3 –9.7 –5.4 –6.3 0.5 0.4 4.7 4.4 –0.7 –3.3 0.5 4.4 –13.1

4 ( p-value = .13)

Source: Uniform Crime Report 1933–2007 Notes: Number of pluses indicates the number of cycles in which growth in the crime rate from trough to peak was less than the rate of growth in the following year.

Fig. 2.3 34



with a tendency to have higher annual rates of change during expansions than during recessions. CZ presented a novel way to make this test more precise by treating each of the business cycles as an experimental trial. A trial is given a plus if the postpeak rate of change exceeds the annual rate of change during the expansion and a minus if the postpeak rate of change in the crime rate is less than the annual rate of change during the expansion. If the business cycle has no impact on crime, then we would expect each result to be equally likely. In this case, each trial can be thought of as a coin toss, in which the coin has a 50% chance of being heads. Since multiple independent coin flips can be modeled using the binomial distribution, the null hypothesis that there is no relationship between the business cycle and crime can be tested using a binomial hypothesis test. In CZ, there were 9 trials, and robbery and burglary both had 8 pluses, an event which would have happened only 2% of the time if there truly was no relationship between crime and the business cycle. Since we have no prior belief about whether the coin should land heads or tails, we actually report a two-tail test: 8 out of 9 with the same outcome occurs only 4% of the time by chance. Auto theft, with 7 out of 9 minuses, appeared procyclical, although this could have happened 18% of the time (two-tailed) even if there was no relationship. As of 2010, we now have 13 trials, which gives us more power to test these same models. The results are presented in table 2.1. Murder has 6 pluses and 7 minuses, which is what we would expect if there were no relationship between homicide and the business cycle. Robbery, on the other hand, had 10 pluses (p-value of 9.2%) and burglary 11 pluses (p-value of 2.2%), suggesting that both are countercyclical. Auto theft is again weakly procyclical, with 4 pluses, which would only happen 13.3% of the time by chance. The US Vital Statistics provide annual time-series data on homicide and suicide victimizations since 1933. These statistics are only available through 2006, so we only have 12 complete cycles. These tests are presented in table 2.1. Both homicide and suicide victimizations lean slightly toward the countercyclical, with 8 out of 12 cycles having higher growth during the recessions than during the expansions. With a p-value of 19.2%, there can be no firm conclusion on this matter.8




_ _ by a first-order autocorrelation process with Eet = 0; Ct , Qt = moving geometric means centered on year t. This specification eliminates secular movements in crime rates caused by trends in crime reporting, demographic structure, “culture,” public policy, and so forth. It also eliminates secular trends in the business conditions indicators, which in recent years have been caused by changes in demographic structure, earlier retirement, the changing role of women in the labor force, and so forth (Leon 1981). This approach accomplishes the same task as the first differencing seen in the work that follows in the tradition of Cantor and Land (1985). In those papers, researchers are attempting to create a stationary series. As discussed in detail by O’Brien (2001), first differencing to achieve stationarity can potentially eliminate the variation of interest to the researcher. In our case, we are interested in preserving the unique characteristics of the business cycle with its roughly five-year boom and then bust cycle while removing secular trends, something that can be achieved with our procedure but not with first differences (Paternoster and Bushway 2001). Appendix A provides a more elaborate justification for the specification. We estimate these regressions using the Prais-Winston correction for serial correlation in the error term. Serial correlation exists when the error term for this period depends on past error terms plus a truly random error term. Prais-Winston assumes the error is first-order autoregressive and then uses generalized least-squares to estimate a transformed model.

Regression Results for Postwar Crime and Violence Rates All regressions were run on data for the period between 1948 and 2008. Because we use the three-year geometric moving averages, the actual analysis begins in 1949 and ends in 2007. For that period, the three indicators of economic conditions are highly correlated with each other.9 The correlation of GDP/capita (in ratio to the three-year geometric mean) with unemployment rate (also in ratio) is – .8507, and with employment it is .8211. The correlation between unemployment and the employment ratio is – .9243. Equation (1) was estimated for the four types of crime (murder,




the robbery equation is .166. It was .17 in the CZ model. The coefficient for employment ratio and robbery is – 2.443, and CZ report a coefficient of – 2.79. The moderate magnitude of the elasticity with respect to unemployment is consistent with prior findings (Bushway and Reuter 2001). In this case, we estimate that a one-percentage-point increase in unemployment from 6% to 7% would increase robbery by 2.8%. The story for auto theft is somewhat more complicated. Auto theft is significantly related to the unemployment rate and employment ratio but not GDP/capita. For each indicator, the magnitude of the point estimate is less in absolute value than that for burglary and in the opposite direction, suggesting that there is a weak procyclical relationship between the business cycle and auto theft. For example, an increase in unemployment from 6% to 7% would decrease auto theft by 1%. Finally, the regression analysis reveals a consistent countercyclical relationship between suicides and all three measures of economic conditions, a relationship that was not evident in the quasi-experimental results. For a more complete treatment of this relationship with similar data, see Luo, Florence, and Quispe-Agnoli (2010). Which of the three indicators is most highly correlated with the measures of crime and violence? In other words, which provides the best prediction for contemporaneous changes in crime? As it turns out, there is no clear winner. GDP/capita has the highest t-statistic for robbery and burglary, while the employment ratio does for auto theft and the unemployment rate for suicide.

Violence against Women While our findings indicate that economic fluctuations have not had much effect on overall homicide rates, it is possible that they have had an effect on particular subsets of this diverse phenomenon. In particular, we might expect that domestic violence and related homicide rates would be influenced by the business cycle. Economic conditions have pronounced effects on household composition, family finances, child care arrangements, and substance abuse rates, each of which could affect the likelihood of serious domestic violence through plausible causal mechanisms. Domestic homicide rates can be estimated directly from the FBI’s




2.5 reports regression results that mimic the approach in table 2.3, but now with the violence variable broken down into three age groups. We see that short-term variation in economic conditions have little effect on homicide victimization rates for people under 15 or over 24. One of our indicators (GDP/capita) has a strong negative association with homicide victimization for the group aged 15 – 24, and the estimated coefficients for the other indicators are suggestive of this countercyclical effect (although not significant). That anomalous finding bears further scrutiny. Since most killers of victims in this age range are also in this age range, the potential explanations incorporate processes that lead to either victimization or perpetration. The suicide results are interesting, since they strongly suggest that while the overall rate is countercyclical (table 2.3), that is not true for youths under 25 — and indeed, we find evidence that for those under 15, suicide is actually procyclical.12 A possible mechanism in this case is guardianship. If adults spend more time with their adolescent children

Table 2.5. Effects of Short-Term Fluctuations in Economic Activity on Homicide and Suicide Victimization, by Age Groups. Estimated Coefficients from Log-Linear Simple Regressions

Homicide: 1948–2006 Age < 15 Age 15–24 Age > 24 Suicide: 1950–2006 Age < 15 Age 15–24 Age > 24

Unemployment rate (S.E.)

Employment ratio (S.E.)

GDP/capita (S.E.)

–.017 (.042) .063 (.052) .003 (.039)

.436 (.606) –.916 (.761) .018 (.576)

.415 (.365) –1.267a (.434) –.420 (.338)

–.313a (.107) .024 (.031) .067a (.018)

3.226c (1.540) –.401 (.424) –.911a (.260)

1.177 (.991) –.104 (.269) –.439b (.168)

Notes: Each coefficient is from a different regression. All variables are in natural log form, so coefficients are elasticities. Each variable is entered as a ratio to its three-year moving geometric average. Estimation method: Generalized least squares with Prais-Winsten correction for first-order serial correlation. a Significant at .01 level b Significant at .05 level c Significant at .10 level


24 Property crime, 1969–2008 All ages Age < 18 Age 18–24 Age > 24 Violent crime, 1969–2008 All ages Age < 18 Age 18–24 Age > 24

Unemployment rate (S.E.)

Employment ratio (S.E.)

GDP/capita (S.E.)

.070c (.039) .010 (.050) .077c (.045) .091b (.041)

–.497 (.483) –.296 (.603) –.425 (.560) –.852 (.508)

–.848a (.254) –.522 (.340) –1.052a (.285) –.862a (.277)

.116c (.064) .027 (.073) .212a (.076) .171a (.065)

–.784 (.775) .334 (.866) –1.844c (.948) –1.637b (.792)

–1.502a (.383) –.932c (.480) –1.984a (.483) –1.836a (.388)

.041 (.063) .030 (.075) .045 (.073) .035 (.058)

.026 (.753) .232 (.881) .058 (.867) –.006 (.691)

–.847b (.423) –.741 (.500) –1.038b (.485) –.680c (.395)

Unemployment rate (S.E.)

Employment ratio (S.E.)

GDP/capita (S.E.)

.082 (.078) .089 (.129) .072 (.094) .218 (.135)

–.722 (.958) –.146 (1.580) –.691 (1.150) –2.513 (1.649)

–1.237b (.515) –1.354 (.867) –1.357b (.628) –1.239b (.517)

B. Specific Crime

Murder All ages Age < 18 Age 18–24 Age > 24

(continued )

>> 43


24 Burglary All ages Age < 18 Age 18–24 Age > 24 Auto theft All ages Age < 18 Age 18–24 Age > 24

Unemployment rate (S.E.)

Employment ratio (S.E.)

GDP/capita (S.E.)

.095 (.084) .034 (.106) .107 (.088) .129 (.081)

–.571 (1.038) –.192 (1.286) –.542 (1.088) –1.177 (.997)

–1.464b (.566) –.965 (.728) –1.710a (.580) –1.484b (.556)

.138b (.058) .063 (.067) .236a (.065) .178a (.059)

–1.008 (.743) –.128 (.828) –2.111b (.847) –1.538b (.751)

–1.531a (.363) –.916b (.448) –2.308a (.382) –1.730a (.377)

–.116c (.068) –.303a (.079) –.011 (.103) .006 (.080)

2.024b (.798) 4.561a (.840) .657 (1.246) .471 (.982)

.051 (.479) 1.791a (.593) –.757 (.726) –.613 (.555)

Notes: Each coefficient is from a different regression. All variables are in natural log form, so coefficients are elasticities. Each variable is entered as a ratio to its three-year moving geometric average. Estimation method: Generalized least squares with Prais-Winston correction for first-order serial correlation. a Significant at .01 level b Significant at .05 level c Significant at .10 level

three age groups tend to have a similar relationship to economic conditions for each of the crime types. The exception, and perhaps the most striking result from table 2.6, is auto theft. For that crime, arrest rates for juveniles are strongly procyclical, while arrest rates for the older groups are not influenced by economic conditions. This result, which we find in all three economic indicators, strongly supports the earlier suggestion by Paternoster and Bushway (2001) that the form of motorvehicle theft associated with youths, joyriding, has a different etiology than thefts by older offenders and is procyclical.

net effect of the business cycle on crime and violence



A final observation involves the findings for robbery and burglary for youths under the age 18. In all cases, the coefficients for arrestees under the age of 18 are substantively (if not significantly) smaller than the coefficients for older arrestees. This pattern achieves significance in the burglary regressions, where the coefficients for 18-year-olds are a fraction of the coefficients for the next-older age group. We found a similar pattern in results for all property crime (Part I) and forgery and fraud (Part II) arrests.15 Although more work is clearly warranted on this topic, a preliminary conclusion is than juvenile arrest rates for property crimes are less sensitive to economic conditions than are arrest rates for adults. Whether that is an accurate reflection of offender behavior (rather than police behavior) is an open question.

Concluding Remarks We set a modest goal for this analysis — determining the net effect of fluctuations in economic conditions on crime and violence. Other analysts have set a more ambitious agenda, going beyond estimation of the net effect to unpack the various causal processes that are brought to bear by the business cycle. The cost of ambition in this case is the loss of clarity, given that the various processes are highly correlated and difficult to distinguish on the basis of available indicators. In this analysis, we have followed CZ in treating the business cycle, and short-term fluctuations in economic conditions, as a black box. Variations in the macroeconomy are inputs to the black box, and variations in the crime rate are outputs. In pursuit of our modest goal, we have analyzed four indicators for that “input” (the NBER cycle dates, the unemployment rate, the employment ratio, and real GDP/capita) and six “outputs” (homicide, suicide, and four Part I crime types). Our measures of crime and violence include the Vital Statistics public health data on suicide and homicide and the UCR data on crimes known to the police and on arrests. We provide a separate analysis of intimatepartner violence, as proxied by the femicide rate. We have also analyzed both Vital Statistics and UCR arrest data by age group to determine if the black box affects youths differently than adults. Table 2.7 provides one simple summary of our results for UCR crime rates. We find very consistent results for robbery and burglary




enter both crime and economic variables as ratios with the three-year moving average, assuring that the coefficient estimates are identified off of short-term variation from local trends. It is entirely possible — really inevitable — that secular changes in economic activity, including changes in technology, the standard of living, demography, and the labor market, influence the nature and volume of crime and crime control. Those secular changes are not our topic, and we warn against extrapolating our findings regarding GDP/capita. For example, it is quite possible that secular growth in GDP will have an effect on homicide rates, even though in the short term there is little or no effect. In addition to analyzing the effect of economic conditions on overall crime and violence rates, we also explore the possibility that youths are affected differently than adults. The clearest example of this disparity is for suicide, for which we have data of high quality and a consistent set of results — namely, that the countercyclical effect of suicide is confined to adults over 24 year old and that the suicide rate for youths under 15 actually appears to be procyclical. The other example of interest is auto theft, which we analyze by use of a shorter time series on arrest rates. What we learn is that the procyclical pattern in auto theft (established from data on crimes known to the police) appears to be confined to juveniles under 18. Since juveniles have traditionally stolen vehicles for joyriding rather than for sale, it is something of a surprise that they (but not adults) are influenced by the economy. It is a reminder once again that variation in economic conditions conveys the possibility of a variety of causal mechanisms, which in some cases may cancel out. For robbery and burglary, the elasticity estimates from age-grouped arrest data are much smaller for juveniles than for adults for all three economic indicators. That result could be accepted at face value, suggesting, for example, that reduced job opportunities during recessions have less effect on juveniles than on adults for these particular crimes. But we would warn not to take that sort of speculation too seriously, given the complexity of the relevant processes. It is also true that we are skeptical of the validity of arrest data as a measure of criminal offending over time. All of the robbery arrest results with respect to the unemployment rate and the employment ratio are null, despite the very strong effects estimated from the series on robberies known to police. That disparity reflects the low correlation between the “crimes known”




Appendix B: Vital Statistics Homicide and Suicide Movements over 10 Business Cycles, by Age Groups A. Homicide Age < 15

Trough-to-peak interval used in calculations

Annual growth trough to peak (%)

Next year (%)

1946–1948 1949–1953 1954–1957 1958–1960 1961–1969 1971–1973 1975–1979 1980–1981 1983–1990 1991–2000

–0.004 –0.047 –0.020 0.087 0.051 0.086 0.023 0.066 0.023 –0.026

–0.090 0.086 0.135 0.109 –0.135 0.035 0.098 0.033 0.067 –0.004

Number of pluses

Age 15–24

+ + + +

+ + +

6 ( p-value = .75)

Annual growth trough to peak (%) 0.024 –0.006 0.018 0.064 0.056 0.004 0.001 0.066 0.043 –0.037

Next year (%) –0.033 0.091 0.091 0.003 –0.043 –0.018 0.042 0.282 0.115 –0.025

Age > 24

+ + +

+ + + +

6 ( p-value = .75)

Annual growth trough to peak (%)

Next year (%)

–0.028 –0.021 –0.007 0.018 0.049 0.045 –0.012 0.066 –0.014 –0.052

–0.065 –0.001 0.004 0.001 0.077 –0.018 0.045 –0.070 0.035 0.278

+ + + + + + +

6 ( p-value = .75)

B. Suicide

Trough-to-peak interval used in calculations 1946–1948 1949–1953 1954–1957 1958–1960 1961–1969 1971–1973 1975–1979 1980–1981 1983–1990 1991–2000 Number of pluses

Age < 15 Annual growth trough Next to peak year (%) (%) –0.183 0.136 0.194 0.064 0.039 0.096 0.000 0.192 0.036 0.010

0.157 –0.389 0.089 –0.211 –0.027 0.216 –0.051 0.207 –0.010 –0.092

3 ( p-value = .34)

+ +

+ +

Age 15–24 Annual growth trough Next to peak year (%) (%) –0.027 –0.020 –0.003 0.034 0.050 0.068 0.018 –0.008 0.016 –0.027

–0.041 –0.045 0.185 –0.011 0.105 0.025 –0.003 –0.014 –0.011 –0.024

3 ( p-value = .34)


+ +


Age > 24 Annual growth trough Next to peak year (%) (%) –0.006 –0.037 –0.006 –0.001 0.001 –0.010 –0.023 0.017 –0.001 –0.017

0.027 0.019 0.096 –0.011 0.024 –0.002 –0.035 0.014 –0.013 0.041

6 ( p-value = .75)

Source: Vital Statistics 1946–2000 Notes: Number of pluses indicates the number of cycles in which growth in the crime rate from trough to peak was less than the rate of growth in the following year.

+ + + + + +





data for homicides are available from 1940 to 2006, and for suicides, from 1950 to 2006. Arrest data are also less reliable because police agencies are less likely to report to the FBI arrests than “crimes known” (Maltz 1999). Moreover, agencies report less consistently than for the overall crime rates, so annual trends for the United States will be based on a shifting pool of reporting agencies. Arrests for robbery appear to be less correlated with offenses reported to the police (r = .36) than for burglary (r = .62). It is possible that the business cycle has more of an effect on police behavior with respect to robbery than to burglary (or vice versa). These results are not reported in this chapter but are available from the authors. In fact, the trend appears to have been sharply downward, with the net result that crime rates have actually decreased between 2007 and 2009. Q represents the employment ratio, the unemployment rate, or the per capita real GDP, none of which is “an intertemporally consistent indicator of economic conditions” due to secular trends in the demographic composition of the labor force and other factors. For this reason, Q is not a good_indicator for B, which is intertemporally consistent by definition. However, (Q/Q ) has a much_stronger claim to being intertemporally consistent and a good indicator for B/B .

References Arvanites, Thomas M., and Robert H. Defina. 2006. Business cycles and street crime. Criminology 44:139 – 164. Bonger, William A. 1916. Criminality and economic conditions. Boston: Little, Brown. Bushway, Shawn, and Peter Reuter. 2001. Labor markets and crime. In Joan Petersilia and James Q. Wilson (eds.), Crime, 3rd ed. Oakland, CA: ICS Press. Cantor, David, and Kenneth C. Land. 1985. Unemployment and crime rates in the post – World War II United States: A theoretical and empirical analysis. American Sociological Review 50 (3): 317 – 332. Cook, Philip J. 2007. Paying the tab: The costs and benefits of alcohol control. Princeton: Princeton University Press. Cook, Philip J., and John MacDonald. 2010. Public safety through private action: An economic assessment of BIDs, locks, and citizen cooperation. Cambridge, MA: NBER Working Paper 15877. Cook, Philip J., and Gary A. Zarkin. 1985. Crime and the business cycle. Journal of Legal Studies 14 (Jan.): 115 – 128. Edberg, Mark, Martha Yeide, and Rick Rosenfeld. 2010. Macroeconomic factors and youth violence: A framework for understanding the linkages and review of available literature. Prepared for the Centers for Disease Control. Gelper, Sarah, Aurelie Lemmens, and Christophe Croux. 2007. Consumer sentiment and consumer spending: Decomposing the Granger causal relationship in the time domain. Applied Economics 39 (1): 1 – 11.


> 53




rates simply to rise when economic conditions sour and to fall when they improve? Or are the consequences for crime of adverse economic conditions contingent on other prevailing factors, yielding increases in crime in some instances but not others? Economic perspectives point to rational assessments of opportunity costs associated with legitimate versus illegitimate economic pursuits and imply that a decline in legitimate income-generating opportunities is likely to yield increases in illegitimate income acquisition, such as robbery, burglary, and theft. However, as typically expressed in the research literature, economic perspectives may not be helpful in explaining why significant declines in the legitimate economy do not invariably produce the anticipated rise in instrumental criminal activity. This is a crucial empirical as well as theoretical issue, because the accumulated body of evidence on the relationship between crime rates and adverse economic conditions (e.g., rising unemployment and falling wages, GDP, and consumer optimism) suggests that sometimes significant downturns in the economy yield an increase in crime and sometimes they do not (e.g., Bushway, Cook, and Phillips, 2010; Chiricos and DeLone, 1992; Cook and Zarkin, 1985; Smith, Devine and Sheley, 1992). Indeed, a cursory glance at figure 3.1 reveals such a pattern for a common indicator of economic adversity: the unemployment rate. The mixed or ambiguous empirical literature on macroeconomic conditions and crime rates sometimes is written off as a function of empirical misspecification (Greenberg, 2001; Raphael and WinterEbmer, 2001). However, another possibility is that adverse economic circumstances yield increases in crime (and good times yield decreases in crime) only under certain conditions. Cantor and Land (1985) drew attention to this idea, noting that increases in some adverse economic conditions, such as unemployment rates, may yield higher crime rates only to the extent that significant simultaneous or subsequent shifts in routine activities do not limit criminal opportunities. Yet several other plausible contingencies are also implied in the literature. For example, as we elaborate later in this chapter, the criminogenic consequences of the most often studied indicators of growing economic adversity — rising unemployment rates, falling wages, depressed economic output, growing consumer pessimism — may be conditioned by the presence of other economic conditions such as the level of inflation, by offsetting




effects on crime may differ depending on rates of inflation, unemployment insurance, illicit drug involvement, incarceration, and the size of the police force. We begin by outlining the theoretical rationales for our study and reviewing the relevant prior research. We then describe our data and methods and present a series of regression models that address our key research issues. We close with a discussion of the implications of our findings for crime theory, research, and policy.

Background and Research Questions Although classic economic arguments about the rational assessment of perceived costs and benefits of legal versus illegal activities understandably represent the modal framework for linking economic adversity and increasing crime (Becker, 1968; Ehrlich, 1973; see also Bushway et al., 2010), several other perspectives are germane as well. Over time, economic hardships can increase crime and violence by fueling community disorder, reducing collective regulation of public spaces, and discouraging spending on the criminal justice system (Wilson, 1996; Edberg, Yeide, and Rosenfeld, 2010). Deteriorating economic conditions also can push low-income consumers into underground markets for stolen goods and may stimulate property crimes in response to growing demand (Rosenfeld and Fornango, 2007). Both processes have been linked to increases in lethal violence (Rosenfeld, 2009). Finally, the stress and frustration associated with job and income loss may lead to crime and violence as coping mechanisms (e.g., Agnew, 1999). The rich theoretical landscape on possible connections between crime and the economy also reveals several possible contingencies that warrant attention when considering whether growing economic adversity is likely to result in increased rates of crime through one or more of the mechanisms just noted. Consideration of these contingencies has significant implications for the conclusions we draw from prior empirical studies, which almost invariably assume a linear, additive effect of economic conditions on crime rates. One potentially important contingency in the link between changes in crime rates and changes in commonly studied economic conditions, such as unemployment rates and wages, is the level of inflation. Rising prices in legitimate markets may enhance the attractiveness of cheap




findings on the relationship between changes in macroeconomic conditions and changes in crime rates. One potentially relevant factor — the extension of unemployment insurance benefits — reflects government fiscal policy regarding how best to respond to economic crises, with an emphasis on easing the adversity of those who lose jobs and earnings. In theory, this type of government social spending can help to temper the criminogenic consequences of significant economic contractions by buffering families and individuals from the full brunt of market forces. This proposition is implied by Messner and Rosenfeld’s (2007) institutional-anomie theory. The theory locates the genesis of motivations for crime in America’s pronounced and universalistic cultural emphasis on the accumulation of wealth, in combination with differential access to legitimate economic opportunities (Merton, 1938). According to the authors, social institutions play a vital role in buffering citizens from the pressures that arise when the cultural and social structures are misaligned in this fashion. A key role of the polity is to provide protection, through income replacement and other means, from the full force of the free-market economic system, especially during significant economic downturns. The provision of unemployment insurance benefits is a common way in which the U.S. government has attempted to soften the blow of significant economic contractions. As McMurrer and Chasanov explain, “The Federal-State unemployment insurance (UI) system, created in 1935, was designed to provide temporary wage replacement for unemployed workers .  .  . and to assist in stabilizing the national economy during cyclical economic downturns” (1995, 30). Some research has shown that government financial assistance can help to reduce crime rates and to reduce the criminogenic effects of high rates of poverty and inequality (e.g., Hannon and DeFronzo, 1998; for a review, see Messner and Rosenfeld, 2007), but we are aware of no prior research on the specific impact of unemployment insurance benefits on crime trends. Significant variation exists in the provision of unemployment insurance benefits during the past three decades, with historically low levels of support provided during the two recessions of the early 1980s and a much greater support during the early 1990s and the 2000s (McMurrer and Chasanov, 1995). We explore the potential implications of this variation in the present research by examining whether the effects on crime




drop during such times (Bushway et al., 2010; Landers, 2008), others have emphasized the expansion of illicit drug economies that can occur when large segments of the population confront significant obstacles in the legitimate economy (Anderson, 1999). Considerable evidence supports the connection between criminal violence and the emergence and proliferation of illicit drug markets (Baumer et al., 1998; Blumstein, 1995; Cork, 1999; Fryer et al., 2006; Goldstein, 1985; Messner et al., 2005; Ousey and Lee, 2007). Less frequently acknowledged, however, is the possibility that the presence of vibrant illegal drug markets can amplify the criminogenic consequences of a significant economic decline by serving as an arena in which increasing numbers of persons swept aside by the contraction of the legitimate economy become embroiled in activities that fuel instrumental property crimes and violence that are part and parcel to many drug markets. This suggests that the association between common indicators of economic adversity and crime rates may be especially strong when drug market activity is relatively high. We assess the merits of this argument by testing for statistical interactions between changes in illicit drug involvement and changes in unemployment rates, wages, GDP, and consumer pessimism. In summary, a broad reading of the literature suggests that anticipating only main effects of economic adversity on crime rates may be overly simplistic. The link between macroeconomic conditions and crime rates may instead be highly contingent on the presence or absence of other factors, including levels of inflation, unemployment insurance benefits, police presence, incarceration, and drug market activity. Prior research rarely has assessed these possible conditional effects of adverse economic conditions. Instead, with some exceptions, the existing research has focused primarily on examining the main effects of economic conditions on crime rates,1 and it has left a trail of mixed and inconclusive results. There is evidence that property crimes, including violent property crimes such as robbery, tend to rise during economic downturns and fall during periods of recovery (for a review, see Blumstein and Rosenfeld, 2008). Also, some prior research indicates that youth unemployment and wage levels are related to youth crime trends in the same fashion: adverse conditions drive crime rates upward (Gould, Weinberg, and Mustard 2002; Grogger 1998). But other studies

Table 3.1. Description of Variables Included in Analysis of Economic Conditions and Crime Rates (N = 82) Variable Dependent variables Logged homicide rate Logged youth homicide rate Logged young adult homicide rate Logged adult homicide rate Property crime rate Explanatory variables City unemployment ratet–1 County average real wage ratet–1 State real GDPt–1 Regional index of consumer sentimentt–1 Regional inflation ratet–1 County per capita unemployment insurance t–1 County per capita income maintenance t–1 City drug involvement ratet–1 City police force sizet–1 State incarceration ratet–1 City % age 15–24 City % black

Variable definition and source Homicides per 100,000 residents, logged (SHR) Homicides involving a victim under 18 per 100,000 residents under 18, logged (SHR) Homicides involving a victim age 18 to 24 per 100,000 residents age 18 to 24, logged (SHR) Homicides involving a victim age 25 and older per 100,000 residents age 25 and older, logged (SHR) Robberies and burglaries per 100,000 persons (UCR) % of civilian labor force unemployed (BLS) Mean real annual wage across all industries (BEA) State-level real gross domestic product (GDP) across all industries (BEA) Summary index of consumer confidence and expectations (Reuters–University of Michigan) Percentage change in regional consumer price index (BEA) Unemployment insurance compensation in real 2008 dollars (BEA) Income maintenance benefits (SSI, SNAP, Family Assistance) in real 2008 dollars (BEA) Residual drug arrest rate obtained after partialling out total arrest rates (UCR) Police officers per 100,000 residents (UCR) Persons incarcerated per 100,000 residents (BJS) % of population ages 15 to 24 (Census Bureau & SEER) % of population who are black (Census Bureau & SEER)


Overall SD

Within-city SD

2.155 .491

.897 2.041

.423 1.672










6.260 41940.95 411915.70

2.633 7760.85 426096.50

1.583 4146.99 257195.90




4.098 133.807 526.369

2.831 91.3568 267.303

2.829 65.240 119.488




202.480 342.620 16.854 19.641

83.823 167.045 3.582 17.631

24.410 142.598 2.120 2.815

Note: UCR = Uniform Crime Reporting Program; SHR = Supplementary Homicide Reports; BJS = Bureau of Justice Statistics; BEA = Bureau of Economic Analysis; BLS = Bureau of Labor Statistics; SEER = Surveillance Epidemiology and End Results; Reuters/University of Michigan = Reuters–University of Michigan Surveys of Consumers (

criminogenic consequences of economic downturns



yield the opposite conclusion (see Chiricos and DeLone, 1992; Baumer, 2008). The ambiguity in findings across studies in the role of adverse economic conditions may reflect differences in model specification or theoretical interpretations (e.g., Cantor and Land, 1985; Greenberg, 2001; Phillips and Greenberg, 2008). Another possibility, though, is that the effects on crime rates of such conditions are highly conditional, an issue we explore in depth later in the chapter.

The Present Study We examine the main and interactive effects on crime rates of several economic conditions for a sample of 82 U.S. cities from 1980 to 2009. We integrated information from a comprehensive city-level database that contains annual indicators of a variety of relevant attributes for more than 100 large U.S. cities (those with 100,000 or more persons based on 2000 population figures) from 1980 to 2004 (Baumer, 2008) with more recent data from a variety of sources. Given the potential significance of the current economic downturn, we gathered data on crime counts for 2009 (such data were not publicly available as this study was under way) for as many of these cities as possible. Overall, we were able to obtain detailed data on homicide disaggregated by age, along with several key explanatory and control measures for 82 large cities. The average population of these cities in 2000 was about 375,000, ranging from just over 100,000 (Livonia City, Michigan) to just over eight million (New York City). Given the time-series panel nature of our data, the analysis is based on 2,460 observations (82 cities × 30 years per city). Data and Methods Table 3.1 lists the variables used in our analysis, providing definitions, sources, and descriptive statistics. Our key dependent variables represent rates (per 100,000 city residents) of property crime (robberies and burglaries), of total homicide, and of homicide disaggregated by age (under 18 years old, 18 to 24, and 25 and older). These measures were constructed using procedures described in appendix A, where we provide additional details about a variety of technical issues associated




ment, incarceration rates, and police force size. Pertinent statistical details of the models are elaborated in appendix A. Results The results of the baseline regression models are displayed in table 3.2. To better illustrate the magnitude of the observed relationships, we present the effects in the form of elasticities. Model 1 shows the findings obtained for property crime rates, and models 2 – 5 show the results for

Table 3.2. Two-Way Fixed Effects Models of the Main Effects of Economic Conditions on Crime Rates (1980–2009, N = 82)

Explanatory variable

Logged Logged young Logged Property Logged youth adult adult crime homicide homicide homicide homicide (1) (2) (3) (4) (5)

Property crime rate

Drug involvement

–.011 (.017) –.033 (.028) –.034 (.100) –.070* (.019) .051 (.191) –.007 (.033) .021 (.011) .041 (.050) –.091* (.044) –.120* (.046) –.022 (.051) .228* (.054)

.275* (.042) .148* (.033) –.011 (.053) –.382* (.187) –.112* (.028) –.936* (.473) –.007 (.091) –.044 (.024) .052 (.073) –.128 (.082) –.338* (.076) .219* (.078) .190* (.070)



City unemployment ratet–1 County average real wage ratet–1 State real GDPt–1 Regional index of consumer sentimentt–1 Regional inflation ratet–1 County per capita unemployment insurance t–1 County per capita income maintenance t–1 City police force sizet–1 State incarceration ratet–1 City % age 15–24 City % black R–Squared * p < .05

.468* (.136) .396* (.144) –.006 (.201) –2.326* (.543) .053 (.071) –5.374* (1.584) –.143 (.335) –.076 (.094) .057 (.303) –.183 (.316) .057 (.215) –.077 (.426) .186 (.341) .327

.458* (.157) .032 (.137) .415* (.203) 1.470* (.674) –.196* (.092) –1.119 (1.467) .199 (.273) –.114 (.096) –.733* (.272) –.613* (.277) –.354 (.249) .780 (.435) .599 (.395) .410

.285* (.057) .171* (.051) –.052 (.087) –.108 (.212) –.195* (.046) –.976 (.651) .129 (.129) –.052 (.049) .122 (.135) –.263* (.131) –.273* (.094) .010 (.135) .193* (.092) .628




we observe that young adult homicide rates are significantly higher in contexts of higher unemployment, an important result given the comparatively high rates of homicide offending and victimization in this age group. This finding also demonstrates the utility of considering agedisaggregated homicide indicators, for if we were to look only at overall homicide rates, our conclusion would be that shifts in unemployment rates are not highly consequential for changes in lethal violence. Finally, we observe no evidence in these models of main effects that shifts in regional inflation rates are significantly associated with changes in crime rates, net of the other factors. We now turn to the central question addressed in our study: are the effects on crime-rate trends of common indicators of economic adversity conditioned by other factors? We assessed this issue by reestimating the equations that form the basis of the results displayed in table 3.2 with the relevant interaction terms included. The resulting unstandardized regression coefficients and standard errors for the multiplicative models estimated in our study are presented in appendix B, with each table focusing on a different moderator variable (e.g., table 3.B1 shows results for the moderating role of inflation levels, 3.B2 shows results for the moderating role of unemployment insurance benefits, and so on). To more clearly capture the nature and magnitude of these results, we summarize the key patterns here by reference to a series of figures, which show the effects of selected economic conditions on crime rates at different levels of other factors. Further details about the model estimation procedures and the construction of the figures can be found in appendix A. Overall, three story lines emerge from this portion of our analysis. First, while the influence of GDP and consumer sentiment on property crime and homicide appear to be invariant across levels of inflation, the effects of city unemployment rates and county wage levels depend on inflation levels in some instances. Figure 3.2 shows at different levels of regional inflation the estimated slopes for the effects of unemployment rates on property crime (panel A) and young adult (ages 18 to 24) homicide rates (panel B) and the estimated slopes for the effects of wage levels on overall homicide (panel C) and adult homicide (panel D). As panel A reveals, the effect of unemployment rates on property crime rates is highly contingent on levels of inflation in a manner consistent




economic downturn. An exception to this pattern, however, is displayed in panel B. Although inflation amplifies the degree to which rising unemployment rates yield increases in property crime, our estimations suggest that it reduces the effect of unemployment on homicide among young adults. The precise meaning of this pattern is unclear. It may be an artifact of measurement error or other features of our approach, but it is also possible that high inflation rates modify the routine activities of young adults in such a way as to reduce their exposure to situations conducive to lethal violence. A second pattern that emerges from our results is that the level of objective risk appears to play a fairly important role in shaping how adverse economic conditions influence crime rates, though not always in ways we expected. Figure 3.3 shows that increases in unemployment rates (panel A) and decreases in GDP (panel B) are less likely to yield elevated property crime when the risk of incarceration is especially high. Stated differently, these results suggest that rising unemployment and falling GDP — two stalwarts of major economic downturns — yield increased property crime primarily under conditions of relatively low incarceration rates. This is consistent with the notion that the threat of custodial sanctions may mitigate the criminogenic consequences of adverse economic conditions, at least with respect to property crime and indirectly for homicide through property crime. But a different picture emerges for the direct effects on homicide and for the effects of wages. Contrary to the expectations outlined earlier, decreasing wages do not appear to have a weaker effect on crime when the risk of incarceration is high. In fact, we find just the opposite. As illustrated in panels C and D of figure 3.3, the inverse association between wages and both property crime and adult homicide is stronger when incarceration rates reach very high levels. Moreover, we observe a similar pattern with respect to the moderating role of police force size on the association between wages and property crime (panel E) and on the association between GDP and homicide rates (panel F). Although contrary to our expectations, a possible implication of these patterns is that increases in wages for legitimate work yield larger reductions in crime when imprisonment levels are high because when wages rise, people immersed in illegal conduct might be more apt to discontinue that conduct if the risk of punishment also is relatively steep.




found in only one instance. Upon further investigation, the lone significant moderating effect of unemployment insurance (table 3.B2, model 2) is relatively small in magnitude and exhibits no significant simple slopes in the focal variable across various ranges of unemployment insurance. The statistically significant interaction between economic conditions and levels of drug involvement that emerged (see table 3.B5, model 3) was larger and substantively meaningful, suggesting that declining GDP yields increases in youth homicide only in the presence of high levels of drug involvement. This is consistent with the idea that drug involvement may amplify the criminogenic consequences of adverse economic conditions. However, the general pattern observed is that levels of drug involvement as measured do not significantly condition the effects of economic conditions on crime levels. A measure that more directly captures illegal drug market activity may yield different findings. Discussion Much of the extant research on the link between indicators of economic downturns and crime rates has focused on estimating main effects, typically evaluating the general hypothesis that declining economic conditions yield elevated crime rates. We propose that this was an overly simplistic approach. There are good theoretical reasons to anticipate the effects on crime rates of commonly examined economic factors to be conditioned by the presence or absence of other factors, and ignoring this possibility in prior research may help to explain why the empirical evidence on the role of economic conditions is highly ambiguous. The primary objective of this chapter was to explore whether the effects of economic adversity on crime rates are conditioned by factors such as levels of inflation, unemployment insurance benefits, illegal drug involvement, and objective levels of sanction risk (police presence and incarceration rates). We pursued this objective with data from 1980 to 2009 on property crime, age-disaggregated homicide, economic conditions, and other factors for 82 relatively large U.S. cities and the counties, states, and regions in which they are located. Our analysis yields evidence that several commonly considered indicators of economic adversity do have effects on crime rates that




elevated crime rates. Our results do not bear this out. It may be that unemployment benefits are simply not of sufficient scale to yield the anticipated mitigating benefits. Higher levels of unemployment insurance or other social spending may be more successful in this regard. Indeed, Johnson, Kantor, and Fishback (2008), although they do not consider a moderating role for social spending, present persuasive evidence that the large-scale relief spending associated with the New Deal helped to keep crime rates low during and in the aftermath of the Great Depression. Additionally, Johnson et al. report that “work relief was more effective than direct relief in reducing crime” (2008, 2). Perhaps the partial income replacement provided by contemporary unemployment benefit programs is not the type of social spending that is likely to mitigate the potentially criminogenic consequences of adverse economic circumstances. We also find little evidence that heightened involvement in illegal drug activity amplifies the effects on crime rates of economic conditions. This is inconsistent with our expectation that declining economic conditions may be more likely to elevate crime in the presence of expanding illegal drug markets. As noted earlier, although our measure of involvement in illegal drug activity partials out city differences in police activity, the measure may still lack the precision needed to test this argument satisfactorily. Even under the assumption that one can effectively purge the contribution of city differences in policing from the drug arrest indicator, one problematic aspect of the measure is that it likely captures some combination of city differences in drug use and drug market activity, and these two dimensions of drug activity may respond in divergent ways to an economic downturn. Specifically, while extant theory seems clear in suggesting that more people may turn to illegal drug markets to replace or augment lost employment and income in the legitimate sector — in other words, that illegal drug markets might grow in such periods — good evidence indicates that alcohol consumption declines as budgets tighten during a weakening economy, and the same may be true of illicit drug use. Thus, declining economic conditions may dampen demand for illicit drugs, even as it may increase the pool of drug sellers and grow illegal drug market participation. The latter mechanism would be expected to amplify the possible criminogenic consequences of adverse economic conditions,




of population change during the period in the counties in which our cities are located and applying those rates of change to the observed decennial points to interpolate annual city estimates. Measurement of Illicit Drug Involvement There are no systematic, direct measures of illicit drug use and drug market activity for American cities for the period under review (National Research Council, 2001). In response to this measurement gap, past research has relied on a variety of proxies, including emergency room drug mentions, drug-related mortality rates, arrestee and prisoner self-reports and drug-testing data, and drug arrest rates. Some research suggests that these indicators are strongly associated across cities and over time (Fryer et al. 2006; Rosenfeld and Decker 1999), and drug arrest rates, based in part on this evidence and their greater geographic and temporal coverage, have emerged as the most common measure employed to gauge differential illicit drug involvement across areas and over time (Blumstein, 1995; Ousey and Lee 2004). The expectation when doing so is that crime rates and drug arrest rates will be positively associated because the latter are thought to reflect differential involvement in illicit drug use and marketing. One limitation of this approach, however, is that the correspondence between drug arrest rates and underlying drug market activity and patterns of illicit drug use is imperfect, and in particular, it contains an unknown degree of error associated with city and temporal differences in police behavior. All else being equal, drug arrests likely will be more prevalent where the police are more productive or, in other words, where overall rates of arrest are higher. From a general deterrence punishment framework, higher arrest rates should yield less crime. Thus, in contrast to the logic just described about the anticipated positive association between crime rates and drug arrest rates (owing to an assumed link between drug involvement and criminal behavior), to the extent that higher drug arrest rates reflect more intensive policing, one might expect an inverse (not a positive) association with crime rates. To minimize the possible confounding effects of drug arrest rates (i.e., that they might be positively associated with crime to the extent that they reflect greater drug involvement and that they might be negatively associated with crime


.90), and we use the broader measure because it better captures the full extent of changes in drug involvement over the period covered in our research. Statistical Analysis Details The econometric panel models estimated to generate the results presented in this chapter include fixed effects that control for stable unmeasured city attributes and temporal shocks that are shared across cities (Raphael and Winter-Ebmer, 2001; Worrall and Pratt, 2004). Consistent with previous research, the annual crime data examined in the study exhibited significant serial autocorrelation, which we accounted for by specifying first-order autocorrelation within panels and panel-specific AR(1) coefficients. Finally, preliminary analyses of the data indicated the presence of substantial cross-sectional correlation in disturbances across cities. Failing to account for these features of the data can lead to invalid inferences, so the models shown in appendix B report panelcorrected standard errors, which allow the disturbances to be heteroskedastic and contemporaneously correlated across panels (Wilson and Butler, 2007). The multiplicative models estimated in appendix B (tables 3.B1 – 3.B5) build on the baseline equations displayed in table 3.2, adding to these equations the relevant interaction (i.e., product) terms. For example, to determine whether regional inflation levels moderate the effects on property crime and logged homicide rates (both overall and disaggregated by age) of city unemployment rates, county average wages, state GDP levels, and regional levels of consumer sentiment, we formed product terms between the inflation indicator and each of the other economic attributes and then added these product terms to the full equations shown in table 3.2. To minimize concerns about multicollinearity

criminogenic consequences of economic downturns



and to enhance the interpretability of the interactions estimated in our regression models, each predictor variable hypothesized to form multiplicative relationships was mean centered prior to computing the product terms (Jaccard and Turrisi, 2003). We present in appendix B only the most pertinent results for these models (the parameters for the main effects of the focus and moderator variables and the interaction terms). The models estimated also included all of the control variables and fixed effects (year and city) dummy variables, but to conserve space we omit these results. We computed the simple slopes implied in tables 3.B1 – 3.B5 using procedures outlined by Aiken and West (1991).

Appendix B Table 3.B1. Two-Way Fixed Effects Models of the Moderating Role of Inflation on the Relationship between Economic Conditions and Crime Rates (1980–2009, N = 82)

Explanatory variable City unemployment ratet–1 X Regional inflation ratet–1 County average real wage ratet–1 X Regional inflation ratet–1 State real GDPt–1 X Regional inflation ratet–1 Regional index of consumer sentimentt–1 X Regional inflation ratet–1 City unemployment ratet–1 County average real wage ratet–1 State real GDPt–1 Regional index of consumer sentimentt–1 Regional inflation ratet–1 R-Squared * p < .05

Logged Logged young Logged Property Logged youth adult adult crime homicide homicide homicide homicide (1) (2) (3) (4) (5) 3.343* (1.606) 1.020 (0.582) 4.193 (14.994) –0.877 (1.071) –7.133 (8.541) –2.001 (4.932) –341.980* (87.963) 1.678 (4.231) –9.972 (16.360) .798

–0.001 (0.001) –0.002* (0.001) –0.013 (0.021) 0.001 (0.001) –0.003 (0.008) –0.011* (0.005) –0.313* (0.075) –0.010* (0.005) 0.003 (0.023) .797

–0.001 (0.005) 0.001 (0.002) 0.010 (0.071) –0.003 (0.006) –0.0003 (0.032) –0.054* (0.013) 0.144 (0.227) –0.061* (0.018) –0.045 (0.085) .327

–0.015* (0.005) –0.000 (0.002) –0.125 (0.072) 0.001 (0.004) 0.062* (0.032) 0.043* (0.016) –0.739* (0.252) –0.010 (0.016) 0.054 (0.070) .416

0.001 (0.002) –0.003* (0.001) –0.005 (0.040) 0.001 (0.002) –0.010 (0.014) –0.007 (0.006) –0.509* (0.135) –0.011 (0.007) 0.032 (0.032) .632




Table 3.B3 (continued )

Explanatory variable City unemployment ratet–1 County average real wage ratet–1 State real GDPt–1 Regional index of consumer sentimentt–1 State incarceration ratet–1 R-Squared

Logged Logged young Logged Property Logged youth adult adult crime homicide homicide homicide homicide (1) (2) (3) (4) (5) –13.820 (8.896) –0.368 (5.257) –831.809* (161.053) 1.050 (4.160) –1,386.594* (256.534) .816

0.003 (0.008) –0.010* (0.005) 0.241 (0.139) –0.009 (0.005) –0.492* (0.224) .800

0.016 (0.031) –0.036* (0.013) 0.850* (0.382) –0.060* (0.018) 0.710 (0.655) .332

0.077* (0.031) 0.018 (0.017) 0.739 (0.435) –0.003 (0.016) –0.380 (0.730) .428

–0.007 (0.014) –0.009 (0.007) –0.276 (0.217) –0.011 (0.007) –0.358 (0.326) .626

* p < .05

Table 3.B4. Two-Way Fixed Effects Models of the Moderating Role of Police Force Size on the Relationship between Economic Conditions and Crime Rates (1980–2009, N = 82)

Explanatory variable City unemployment ratet–1 X Police force size–1 County average real wage ratet–1 X Police force sizet–1 State real GDPt–1 X Police force sizet–1 Regional index of consumer sentimentt–1 X Police force sizet–1 City unemployment ratet–1 County average real wage ratet–1 State real GDPt–1 Regional index of consumer sentimentt–1 City police force sizet–1 R-Squared *p < .05

Logged Logged young Logged Property Logged youth adult adult crime homicide homicide homicide homicide (1) (2) (3) (4) (5) –151.516 (79.209) –225.684* (47.554) –989.839 (730.727) –10.950 (11.362) –10.021 (8.391) –0.839 (4.073) –615.588* (104.278) 2.686 (4.383) –722.504 (410.644) .796

–0.085 (0.066) 0.040 (0.033) –1.722* (0.768) –0.012 (0.010) –0.002 (0.008) –0.009 (0.004) –0.373* (0.075) –0.011 (0.006) –0.712 (0.416) .796

–0.069 (0.246) –0.101 (0.121) –3.797 (2.109) –0.003 (0.035) –0.004 (0.032) –0.053* (0.013) –0.211 (0.208) –0.063* (0.018) –1.006 (1.492) .331

–0.124 (0.257) 0.078 (0.116) –6.457* (3.253) –0.061 (0.040) 0.064* (0.032) 0.039* (0.016) –0.972* (0.276) –0.010 (0.017) –2.929* (1.359) .409

–0.075 (0.090) 0.001 (0.040) –1.939* (0.804) 0.004 (0.015) –0.009 (0.014) –0.002 (0.005) –0.608* (0.107) –0.012 (0.008) –1.207 (0.618) .626




Baumer, Eric P. 2008. “An Empirical Assessment of the Contemporary Crime Trends Puzzle: A Modest Step toward a More Comprehensive Research Agenda.” In Arthur Goldberger and Richard Rosenfeld (eds.), Understanding Crime Trends. Washington, DC: National Academies Press. Baumer, Eric P., Janet L. Lauritsen, Richard Rosenfeld, and Richard Wright. 1998. “The Influence of Crack Cocaine on Robbery, Burglary, and Homicide Rates: A Cross-City, Longitudinal Analysis.” Journal of Research in Crime and Delinquency 35:316 – 340. Becker, Gary. 1968. “Crime and Punishment: An Economic Approach.” Journal of Political Economy 73:169 – 217. Blumstein, Alfred. 1995. “Youth Violence, Guns, and the Illicit-Drug Industry.” Journal of Criminal Law and Criminology 86:10 – 36. Blumstein, Alfred, and Richard Rosenfeld. 2008. “Factors Contributing to U. S. Crime Trends.” In Arthur Goldberger and Richard Rosenfeld (eds.), Understanding Crime Trends. Washington, DC: National Academies Press. Bushway, Shawn, Philip J. Cook, and Matthew Phillips. 2010. “The Net Effect of the Business Cycle on Crime and Violence.” Prepared for the Centers for Disease Control. Cantor, David, and Kenneth C. Land. 1985. “Unemployment and Crime Rates in the Post – World War II United States: A Theoretical and Empirical Analysis.” American Sociological Review 50:317 – 332. Chiricos, Ted, and Miriam DeLone. 1992. “Labor Surplus and Punishment: A Review and Assessment of Theory and Evidence.” Social Problems 39:421 – 446. Cook, Philip J., and Gary A. Zarkin. 1985. “Crime and the Business Cycle.” Journal of Legal Studies 14:115 – 128. Cork, Daniel. 1999. “Examining Space-Time Interaction in City-Level Homicide Data: Crack Markets and the Diffusion of Guns among Youth.” Journal of Quantitative Criminology 15:379 – 406. Curtis, Lynn A. 1981. “Inflation, Economic Policy, and the Inner City.” Annals of the American Academy of Political and Social Science 456:46 – 59. Devine, Joel A., Joseph F. Sheley, and M. Dwayne Smith. 1988. “Macroeconomic and Social-Control Policy Influences on Crime Rate Changes, 1948 – 1985.” American Sociological Review 53:407 – 420. Edberg, Mark, Martha Yeide, and Rick Rosenfeld. 2010. “Macroeconomic Factors and Youth Violence: A Framework for Understanding the Linkages and Review of Available Literature.” Prepared for the Centers for Disease Control. Ehrlich, Isaac. 1973. “Participation in Illegitimate Activities: A Theoretical and Empirical Investigation.” Journal of Political Economy 81:521 – 565. FBI. 2010. “FBI Releases 2009 Crime Statistics.” documents/pressreleasecius09.pdf. FBI. 2011a. “FBI Releases 2010 Crime Statistics.” crime-in-the-u.s/2010/crime-in-the-u.s.-2010/summary. FBI. 2011b. “Preliminary Semiannual Uniform Crime Report, January – June, 2011.”




McMurrer, Daniel, and Amy Chasanov. 1995. “Trends in Unemployment Insurance Benefits.” Monthly Labor Review 118:30 – 39. Merton, Robert K. 1938. “Social Structure and Anomie.” American Sociological Review 3:672 – 682. Messner, Steven F., Glenn D. Deane, Luc Anselin, and Benjamin Pearson-Nelson. 2005. “Locating the Vanguard in Rising and Falling Homicide Rates across U.S. Cities.” Criminology 43:661 – 696. Messner, Steven F., and Richard Rosenfeld. 2007. Crime and the American Dream. Belmont, CA: Wadsworth. National Bureau of Economic Research. 2010. “Business Cycles and Expansions.” National Research Council. 2001. Informing America’s Policy on Illegal Drugs. Washington, DC: National Academy Press. Ousey, Graham C., and Matthew R. Lee. 2004. “Investigating the Connections between Race, Illicit Drug Markets, and Lethal Violence, 1984 – 1997.” Journal of Research in Crime and Delinquency 41:352 – 383. Phillips, Julie A., and David F. Greenberg. 2008. “A Comparison of Methods for Analyzing Criminological Panel Data.” Journal of Quantitative Criminology 24:51 – 72. Ralston, Roy W. 1999. “Economy and Race: Interactive Determinants of Property Crime in the United States, 1958 – 1995: Reflections on the Supply of Property Crime.” American Journal of Economics and Sociology 58:405 – 434. Raphael, Stephen, and Rudolf Winter-Ebmer. 2001. “Identifying the Effect of Unemployment on Crime.” Journal of Law and Economics 44:259 – 283. Rosenfeld, Richard. 2009. “Crime Is the Problem: Homicide, Acquisitive Crime, and Economic Conditions.” Journal of Quantitative Criminology 25:287 – 306. Rosenfeld, Richard, and Scott H. Decker. 1999. “Are Arrest Statistics a Valid Measure of Illicit Drug Use?” Justice Quarterly 16:685 – 699. Rosenfeld, Richard, and Robert Fornango. 2007. “The Impact of Economic Conditions on Robbery and Property Crime: The Role of Consumer Sentiment.” Criminology 45:735 – 769. Seals, Alan, and John Nunley. 2007. The Effects of Inflation and Demographic Change on Property Crime: A Structural Time Series Approach. Working Paper, Middle Tennessee State University. Surveillance Epidemiology and End Results. 2010. US Population Data — 1969 – 2009. Smith, M. Dwayne, Joel A. Devine, and Joseph F. Sheley. 1992. “Crime and Unemployment: Effects across Age and Race Categories.” Sociological Perspectives 35:551 – 572. Tang, Chor Foon. 2009. “The Linkages among Inflation, Unemployment and Crime Rates in Malaysia.” International Journal of Economics and Management 3:50 – 61. Tang, Chor Foon, and Hooi Hooi Lean. 2007. “Will Inflation Increase Crime Rate? New Evidence from Bounds and Modified Wald Tests.” Global Crime 8:311 – 323. U.S. Census Bureau. 2010. American Fact Finder. nav/jsf/pages/index.xhtml.


> 85




United States for the period 1973 – 2005 and provide the first detailed examination of how these trends are associated with several indicators of national economic conditions, including consumer pessimism, poverty, and unemployment.

Previous Research Aside from the research presented in this volume, there is relatively little work assessing how changes in economic conditions are associated with changes in youth violence rates. Nonetheless, there is a substantial body of cross-sectional research that has found a relationship between economic disadvantage and violence at various levels of analysis. Generally speaking, a strong and significant relationship between economic conditions and rates of violence has been found in studies of census tracts (e.g., Krivo, Peterson, and Payne, 2009), neighborhoods (e.g., Sampson, Raudenbush, and Earls, 1997), cities (e.g., Land, McCall, and Cohen, 1991), and metropolitan areas and states (e.g., Land, McCall, and Cohen, 1990). A recent meta-analysis of 153 studies has shown that there is a strong and stable relationship between poverty and violence in past research (Pratt and Cullen, 2005). The consistency of these crosssectional findings suggests that changes in economic conditions should be associated with changes in youth violence in analyses of temporal patterns in both phenomena. However, aside from the crime of robbery — an economically motivated form of violence — only a handful of studies have demonstrated a relationship between economic trends and changes in violence. Statelevel changes in unemployment or wages (Raphael and Winter-Ebmer, 2001) and in gross state product (Arvanites and Defina, 2006) have been found to be associated with state-level robbery trends. Rosenfeld (2009) found that changes in regional economic conditions, as measured by the Index of Consumer Sentiment, have an indirect effect on homicide through their effect on economically motivated crimes such as robbery, burglary, and auto theft. Rosenfeld’s findings suggest that economic downturns can indirectly affect homicide rates by increasing the demand for stolen goods through underground or illegal markets, which are supplied by such acquisitive crimes. Increases in underground market activity can prompt increases in homicide because they




examine how changes in indicators of economic deprivation and intraracial inequality affected changes in age-race-specific arrest rates. As measures of economic deprivation, Messner et al. included information on median family income and the percentage of children under 18 years of age living in poverty. Intraracial inequality was measured using the Gini coefficient, the interquartile income range, and the share of income received by the top 5% of households. Messner et al. found that changes in economic deprivation are positively associated with changes in juvenile offending as measured by UCR homicide arrest data. They also found that increases in unemployment had a lagged, negative effect on black and white juvenile arrest rates for homicide, a finding that was unexpected and in contrast with the influence of poverty. Intraracial inequality (but not interracial inequality) was shown to be associated with higher arrest rates for youths; however, these relationships were contingent on the type of measure and lag structure used. Messner et al. note that the complex findings about inequality and unemployment should be regarded with caution. They comment that there are “few clear theoretical guidelines in the literature . . . to govern the selection of specific measures” (2001, 605), and this caveat also is relevant to the findings presented in this chapter. The analysis that follows shows how changes in economic conditions are associated with changes in violent victimization for non-Latino black (hereafter “black”), Latino, and non-Latino white (hereafter “white”) male and female youth ages 12 to 24 in the United States. We begin by describing the data and estimation procedures we use for producing reliable long-term trends in youth violence, as well as the various national economic indicators used in our analysis. Our selection of economic indicators was informed by past research. Preliminary investigations of several sets of possible indicators determined which economic conditions would be the focus of subsequent analysis. Those that were found to be most strongly associated with the trends are displayed in figures to illuminate the nature of these relationships; however, we also include details about conditions that we found not to be related to the trends. Although these analyses are bivariate and descriptive in nature, they are the necessary first step toward understanding whether economic changes are associated with changes in youth violence.




are used to code the incident into crime types following the interview. We define male serious violence to include attempted or completed robberies and aggravated assaults, which are defined as attacks in which the perpetrator used a weapon or those that resulted in injury (Bureau of Justice Statistics, 2005).3 Female serious violence includes attempted or completed rapes in addition to attempted or completed robberies and aggravated assaults. We also assess trends in gun violence for males and females by age and by race and ethnicity. Gun violence represents a subset of serious nonfatal violence and includes any form of violent victimization in which the victim reported that the offender had a gun.4 Race and ethnicity as well as age are measured using self-reports to questions created and used by the Census Bureau. Following Census practices, NCVS items on race and ethnicity have changed over time.5 To create a set of consistent categories, we combine responses to the “race” questions with responses to the “ethnicity” question and code for the three largest race and ethnic groups in the nation: non-Latino blacks, non-Latino whites, and Latinos (persons of Hispanic origin or descent who may be of any race). Unfortunately, there are insufficient numbers of subjects to provide reliable annual estimates of youth violence for other race and ethnic groups. For the purpose of these analyses, the key strengths of the NCVS data are that they permit custom estimations of subgroup rates of victimization, they include incidents not reported to the police, and the data have been available for more than 35 years. Aside from the redesign, there have been no methodological changes that would be confounded with observed changes in violence rates.6 Nonetheless, these data are subject to concerns that affect all survey data, such as recall error and under- and overreporting. Potential sources of error in the NCVS have been studied extensively (Groves and Cork, 2008; Lynch and Addington, 2007; Penick and Owens, 1976), and two issues that may be relevant to the trends are noted here. First, the use of adjustment weights for the pre-1992 data assumes that the effect of the new instrument would have been the same had it been phased in at a different time point; however, available data cannot assess this assumption (Rand, Lynch, and Cantor, 1997). Second, the willingness of respondents to report violence to interviewers may have changed over time. This concern applies to all




the available poverty-trend data, the unemployment-trend data from the CPS do not distinguish non-Latino whites and non-Latino blacks from Latinos. Following research by Rosenfeld and Fornango (2007), we also assess the Index of Consumer Sentiment (ICS).9 The ICS is a summary measure derived from a longstanding and ongoing survey of U.S. adults that includes questions about personal finances, assessments of the economy now and in the short and long term, and consumers’ willingness to purchase large household items. The ICS captures the “subjective experience of economic hardship and change” (Rosenfeld and Fornango, 2007, 740) and is used as part of the Index of Leading Economic Indicators (Conference Board, 2009). According to Curtin (2004), changes in the ICS predict changes in GDP and foreshadow the onset of recessions. In the current analyses, we reverse the scaling of the ICS axis so that the peaks represent greater levels of consumer pessimism. Unfortunately, it is not possible to get ICS trend data disaggregated for nonLatino white, non-Latino black, and Latino subgroups.10 Our choice of economic indicators focused on poverty, unemployment, and the ICS. We examined each of the trends just noted and, for poverty and unemployment, selected the one that was found to be most strongly associated with each of the violence trends to become the focus of our subsequent analysis.11 To represent changes in poverty, we found that the youth poverty measure (i.e., percentage of youth ages 18 and under living below poverty) generally outperformed indicators of the overall poverty rate and the race/ethnic-group-specific poverty rates (when applied to their respective subgroup violence trends). Interestingly, the youth poverty trend also outperformed the race- and ethnic-group-specific youth poverty trends for each of the race- and ethnic-group-specific violence trends. In other words, greater subgroup specificity in economic indicators did not result in higher correlations, a finding consistent with prior analyses (e.g., Messner et al., 2001). We conducted a comparable set of analyses for our choice of unemployment indicator and found that the trend in youth unemployment (i.e., percentage of persons ages 16 to 24 who are unemployed) similarly outperformed the overall unemployment trend as well as the race- and ethnic-group-specific unemployment trends. Thus, the following analysis




Fig. 4.2. Homicide (per 100,000), gun violence (per 10,000), serious violent victimization (per 1,000), and consumer pessimism: males ages 12 – 17

Trends in Homicide, Gun Violence, and Serious Nonfatal Violence by Age and Gender Our first set of youth victimization trends examines the similarities and differences in male and female patterns of nonfatal and fatal violence for younger and older youth and compares these trends to changes in economic conditions. Figure 4.2 displays the trends in homicide, and our estimates of the trends in nonfatal gun violence and serious violent victimization for males ages 12 – 17. In this and all subsequent figures, we display our estimates of the annual rates in nonfatal violence using three-year moving averages to better depict the overall trends, and we use different base rates for the various forms of violence (i.e., per 100,000 for homicide, per 10,000 for gun violence, and per 1,000 for serious violence) to ease comparisons of the similarities and differences in the trends. In figure 4.2, the national trend in consumer pessimism is also superimposed. Changes in the consumer pessimism measure are presented in this figure, as opposed to the youth poverty or youth unemployment measures, because generally speaking, we found that the consumer pessimism measure was more strongly and consistently associated with these male violence trends (see table 4.1).




(approximately 173% and 208%, respectively), rates of nonfatal serious violence were not increasing to the same degree (35%). In 1993, both homicide and gun violence began to decline at nearly the same rates that they increased, and at that time the rates of serious violence also began to decline. The decline in each of the violence trends ended around 2001 – 2002, and from then until 2005, only rates of homicide began to show some increase. Thus, the large increase in homicide for this subgroup during the 1984 – 1993 period was associated with a large increase in gun violence but does not appear to have been driven by a large increase in serious violence more generally. The correspondence between these violence trends and economic conditions is somewhat complex. Increases in consumer pessimism exhibit somewhat similar patterns as increases in violence in the late 1970s and early 1990s, but it appears that the change in economic conditions was associated with a proportionately small change in male youth violence in the earlier period compared to the later period and that the increase in consumer pessimism in 2001 is not matched by a correspondingly large increase in violence from 2001 to 2005. The first-difference correlations between consumer pessimism and homicide (r = .48) and serious violence (r = .33) are statistically significant ( p  ≤  .05, one-tail test) for the 1973 – 2000 period, but the relationship for gun violence (r  =  .23) during this period is not large enough to reach statistical significance (see table 4.1).13 When the correlations are estimated for the full 1973 – 2005 period, we find that only homicide (r = .38) remains significantly associated with ICS for males age 12 – 17 (but see note 13). Thus, we find some evidence that changes in violent victimization among males age 12 – 17 are significantly associated with changes in economic conditions as measured by the consumer pessimism index. The trends in violence for males ages 18 – 24 in figure 4.3 exhibit patterns that are somewhat similar to those found for males ages 12 – 17. For the older males, there is a comparatively small periodic increase and decrease in homicide during the late 1970s to early 1980s, and as with the younger males, gun violence and homicide began to grow around 1984, while rates of nonfatal serious violence were not increasing to the same degree. Around 1993, both homicide and gun violence began to decline at roughly the same time that the rates of serious violence also




The trends in violence among females ages 12 – 17 in figure 4.4 differ considerably from the corresponding trends for males. For younger females, there are comparatively smaller periodic increases and decreases both in nonfatal serious violence and homicide throughout the entire period 1973 – 2005. Serious nonfatal violence and homicide rates were quite steady during the 1970s and until the mid-1980s. Starting in 1983, both rates began to increase and continued to increase until 1993. After that, both serious violence and homicide exhibit steady decreases until 2005. The trends indicate that nonfatal serious violence and homicide trends among females ages 12 – 17 exhibit similar behavior throughout the period of interest, with the homicide rates fluctuating slightly more over the years. Gun violence for females in this age group was increasing slowly but steadily from the late 1970s to around 1990. Compared to the 1984 – 1993 increases in serious violence (57%) and homicide (44%), the increase in gun violence during this period was much higher (160%). From 1990 to 1993, the rate of gun violence exhibited a particularly dramatic increase, when it almost tripled in magnitude. All three violence rates declined steadily starting in 1993, with the exception of a few short-term increases in the homicide rate. It is important to note that compared to young males, the large increases in gun

Fig. 4.4. Homicide (per 100,000), gun violence (per 10,000), serious violent victimization (per 1,000), and youth poverty: females ages 12 – 17




Fig. 4.5. Homicide (per 100,000), gun violence (per 10,000), serious violent victimization (per 1,000), and youth poverty: females ages 18 – 24

tions between violence T and youth unemployment T are generally much weaker than they are between violence T and consumer pessimism T or youth poverty T . Thus, all three economic indicators appear to be capturing important but different aspects of economic conditions, and changes in some of these conditions are associated with serious violent victimization. The trends in violence for older females in figure 4.5 are in many ways similar to the trends for the younger category of females. The trends in serious violence and homicide for the older females are also quite stable across the period, although in absolute terms they exhibit higher fluctuations than for younger females. Serious nonfatal violence for older females was very steady from 1973 to 1990, exhibiting a modest increase from 1990 to 1993, before declining until 2005. Homicide rates for this group of females have been similar to serious violence: very stable until the early 1990s, followed by general declines thereafter. The rates of gun violence have followed a different pattern compared with the other two groups of rates. Gun violence rates in the early 1980s appear to be similar to their levels in the early to mid-1970s, and then they decreased from approximately 1981 to 1986. In 1987, gun violence began to increase quite rapidly and peaked in 1993, similar to the pattern among younger




corresponding economic trend, while youth poverty is included with the female youth violence trends. Recall that measures of poverty and unemployment disaggregated by race and ethnicity (and by age, race, and ethnicity) did not provide a better fit to their respective subgroup trends than did the measures of consumer pessimism or youth poverty. Figure 4.6 shows the trends in serious nonfatal violence for males ages 12 – 24, and table 4.2 displays the first-difference correlations between these trends and our economic indicators. Figure 4.6 shows that minority males experience higher rates of serious violent victimization, particularly during periods of economic downturns in which consumer pessimism is high. During periods in which consumer pessimism is low, the differences in male violence across these race and ethnic groups appear to be much reduced. Thus, what is apparent in these disaggregated violence trends that was not visible in the earlier figures are the differences in levels of risk by race and ethnicity, and the greater periodic fluctuations in the minority trends that coincide with periods of economic downturns. It is also the case that serious violence declined considerably during the 1990s for all three groups. The relationships between consumer pessimism and the serious violence trends for black (r = .52), Latino (r = .39), and white (r = .39)

Fig. 4.6. Serious violent victimization (per 1,000) by race and ethnicity and consumer pessimism: males ages 12 – 24




Figure 4.7 displays the trends in gun violence for males ages 12 – 24, illustrating important differences in level and peaks in this form of victimization across race and ethnic groups. These data show that young black males have consistently faced the highest risks of gun violence over time but that during the early to mid-1990s, Latino rates were nearly as high. In addition, the large increases in gun victimization for black and Latino male youth that appeared around 1984 – 1985 began several years prior to the increases for white male youth, which did not appear to increase until approximately 1989. These trends also show that young minority males experience larger periodic increases in gun violence than do white males during times in which consumer pessimism is increasing, though missing gun data in 1977 and 1978 make it difficult to discern whether this was the case in the late 1970s. When consumer pessimism is low, the patterns are complex. During the mid-1980s, the level differences between blacks, whites, and Latinos in male gun violence were proportionately smaller than they were when consumer pessimism was at its lowest levels during the late 1990s. Nonfatal gun violence declined considerably during the 1990s for all three groups, and like other forms of violence considered here, these declines leveled off beginning in the early 2000s.

Fig. 4.7. Gun violence (per 1,000) by race and ethnicity and consumer pessimism: males ages 12 – 24




Fig. 4.9. Gun violence (per 1,000) by race and ethnicity and youth poverty: females ages 12 – 24

violence declined starting in the middle 1990s for all three race and ethnic groups. Among Latina and black females, the decline leveled off, but it continued for white females into the early 2000s. The relationships between youth poverty and serious violence trends for black (r = .48) and Latina (r = .41) females are statistically significant ( p  ≤  .05, one-tail test) for the 1973 – 2000 period, and these relationships remain significant when the period is extended to include the later years (see table 4.2). However, this relationship is weaker and does not reach significance for white females, regardless of the time-period specification (r = .31 and r = .29). Thus, the findings about the relationship between changes in economic conditions and serious violence for young females by race and ethnicity are generally similar to the overall patterns found for young males. Finally, figure 4.9 shows gun violence against young females ages 12 – 24 by race and ethnicity. The pattern is similar to that for young women’s serious violent victimization. Young black girls and women experience consistently higher rates of gun violence victimization than do young white females and Latinas, although black-Latina differences vary more over time. During the 1980s, there is some increase in gun violence against blacks that corresponds with increases in youth poverty




refined disaggregation by victim characteristics that enables us to present rates of nonfatal violent victimization for Latino/a youth as well as non-Latino/a black and non-Latino/a whites. Our findings suggest that reliance on UCR data can mask important similarities between Latino and black males, as well as important differences between Latino and white males. Moreover, the NCVS data allows us to disentangle nonfatal violence involving guns from serious violence more generally, which allows for more detailed analyses of the relationships between fatal and nonfatal violence and the use of guns in violent crime incidents. As described earlier, our findings reveal many differences and similarities in youths’ exposure to violent victimization over time across gender, race, and ethnicity. Throughout our descriptions, we have examined correspondences between shifts in major indicators of the economy and youth violent victimization, in an attempt to uncover possible connections and to lay the foundation for subsequent research on the macroeconomic contexts of youth violence. First, we observe that there is some association between economic shifts, as measured by consumer pessimism, and trends in youth violence by males, particularly homicide and gun violence. The increases in consumer pessimism in the early 1980s corresponded with modest increases in homicide and gun violence among males ages 12 – 17 and more substantial increases in homicide among males ages 18 – 24. Interestingly, the increases in consumer pessimism in the early 1990s — although at lower absolute levels than in the early 1980s — were associated with much larger upward swings in youth homicide and gun violence among males ages 12 – 17 and much greater increases in homicide among males ages 18 – 24. This suggests that while macroeconomic conditions are associated with male violent victimization, there was something unique during the late 1980s and early 1990s that also helped to drive the strong increases in homicide and gun violence. As suggested in past research, perhaps urban decay and the proliferation of crackcocaine markets that also occurred at this time contributed to and magnified the effects of the broader economy. Certainly, previous research has connected the use of guns and crack markets with violence during this period (Blumstein and Rosenfeld, 1998; Blumstein and Wallman, 2006; Cook and Laub, 2002). Corresponding with the substantial decline in consumer pessimism after 1992, our data show a sharp




Our disaggregation of NCVS trends in youth violence by race and ethnicity, in addition to gender, produces some striking findings that at times parallel recent findings that were not disaggregated by age, published in our other work (Lauritsen and Heimer, 2010). First, young black males ages 12 – 24 have the highest rates of violent victimization over time. During the crime boom period of the late 1980s through early 1990s, however, young male Latino rates of both serious violence and gun violence approached the non-Latino black rates. Second, the rates of gun victimization in both groups of minority male youths show greater correspondence with consumer pessimism during this period than is the case for non-Latino white male youths, even though the rates of serious violence for all three groups are associated with changes in consumer pessimism. Moreover, unlike the trend lines for young males that were not disaggregated by race, serious violence and gun violence against both blacks and Latinos also increased in the early 1980s, again corresponding to trends in consumer pessimism. While we saw a modest association in the trends in violence and consumer pessimism when the data were not distinguished by race and ethnicity, the disaggregated data reveal a pattern that shows that the violent victimization of young minority males coincides somewhat more closely with macroeconomic conditions than is the case for white males. Lauritsen and Heimer (2010) argue that this may occur because minorities tend to be hardest hit during economic downtowns. However, we also note the greater increase in violence against minority males — particularly gun violence — during the 1990s crime boom. As discussed earlier, this may reflect a combination of economic factors, urban decay, and the growth in the crack-cocaine market that occurred during that period. Among young females, by comparison, black females have the highest rates of both serious violence and gun victimization throughout the series, with a clear upward spike during the early 1990s, followed by a dramatic decline. Unlike the patterns for young males, Latina violent victimization is closer in level and pattern over time to white than to black victimization trends, indicating an important interaction between ethnicity and gender. For all three groups of females, fluctuations in economic conditions in the 1980s correspond much more weakly with




The currently available data cannot tell us whether the decoupling in the male series will continue or whether the economic downturn that began in late 2007 will be associated with a subsequent increase in violent victimization. It is almost impossible to assess the cause of this break with the data presented here. It could be that this is to some extent a byproduct of the economic indicator that we use for males. We selected consumer sentiment (consumer pessimism here) for comparison with male victimization rates because it is the indicator most closely associated with violence against young males, whereas we compared trends in youth poverty among females. It may be that the consumer sentiment index was sensitive to the terrorist events of September 11 but that the dot-com economic recession of the early 2000s had relatively little impact on the life circumstances of the young males most exposed to violence; indeed, youth poverty rates increased some from 1999 to 2002 but then leveled off and even decreased slightly. Regardless, the association between consumer sentiment and male victimization became less strong after 2000, and better understanding the reason for this will require additional data. The effect of the current economic crisis on youth victimization cannot be known or predicted with great accuracy. It appears that the economic crises of the 1980s and 1990s may have had different effects on youth violence, perhaps because of other time-bound contextual factors, including the proliferation of crack-cocaine markets, gang activity, and the pace of urban decay. It is also the case that government response varies across periods of economic decline and recession. The recent infusion of funds into government programs (such as extended unemployment benefits) and infrastructure may have had the effect of blunting some of the impact of the most recent recession, or it may be the case that criminal justice policies such as police practices have become more effective at deterring crime. These are important issues for future research. What our analysis offers is evidence that, over a long time series, there is an association between indicators of economic decline and violence against youth. However, careful theoretical development is necessary to understand the specific causal mechanisms underlying these relationships, and more research is needed to specify the precise nature of variation in economic effects across race, ethnicity, and gender.




change period showed that the 1985 estimate of the Latino/Latina population was slightly lower than the 1986 estimate based on the “Hispanic” versus the “nonHispanic” option — a finding consistent with population trends. We found that no adjustments beyond those for type of crime were necessary (see Lauritsen and Heimer, 2010). For rates of serious violent victimization, we weighted NCS incidents of aggravated assault by 1.23 and rape by 2.57 (see Rand, Lynch, and Cantor, 1997). We found no need to adjust the NCS gun violence rates because the redesigned questionnaire did not result in increased reporting of these types of incidents. For 1973 – 1979, we used the SHR data made available through ICPSR (study no. 4351, “Uniform Crime Reports [United States]: Supplementary Homicide Reports 1976 – 2003”). We generated the rest of the series using Easy Access to the FBI’s Supplementary Homicide Reports: 1980 – 2006. For more information, see SHR trends are correlated with the WISQARS estimates at .99 for the years that both series have in common (1981 – 2005). A complete discussion of the ICS is found in Curtin (2004) and Rosenfeld and Fornango (2007). Rosenfeld and Fornango find that consumer sentiment accounted for a significant amount of the change in robbery and property crime independent of the effects of unemployment, GDP, age and race composition, and police and imprisonment. The ICS data are available at The ICS data can be obtained to allow for “black” versus “nonblack” comparisons. Our analyses of the relationships between the black and nonblack ICS trends and the various race- and ethnic-subgroup violence trends showed that the black and nonblack ICS trends did not have a stronger relationship to subgroup trends in violence compared to the aggregate ICS measure. We compared first-difference correlations for each of the trends and selected the economic indicator that produced the largest and most consistent set of associations with the violence trends. The National Bureau of Economic Research (NBER) defines recessions as those periods in which there is a significant decline in economic activity across a variety of indicators including GDP, income, employment, industrial production, and retail sales (NBER, 2003). Recession periods begin when significant decreases in these indicators become evident and end when significant increases resume. Since 1973, the NBER has designated six periods as recessions: November 1973 to March 1975, January 1980 to July 1980, July 1981 to November 1982, July 1990 to March 1991, March 2001 to November 2001, and the most recent recession, which began in December 2007 (NBER, 2010). This may appear odd because the patterns of gun violence and homicide are so similar. We find that for gun violence, the relationship is statistically significant (r = .42 for 1973 – 2000, and r = .45 for 1973 – 2005) if ICS at T – 1, rather than T, is used.




Land, Kenneth C., Patricia L. McCall, and Lawrence E. Cohen. 1991. Characteristics of U.S. cities with extreme (high or low) crime rates: Results of discriminant analyses of 1960, 1970, and 1980 data. Social Indicators Research 24:209 – 231. Lauritsen, Janet L., and Karen Heimer. 2008. Gender and violent victimization, 1973 – 2004. Journal of Quantitative Criminology 24:125 – 147. Lauritsen, Janet L., and Karen Heimer. 2010. Violent victimization among males and economic conditions: The vulnerability of race and ethnic minorities. Criminology & Public Policy 9:665 – 692. Lynch, James P., and Lynn A. Addington, eds. 2007. Understanding crime statistics: Revisiting the divergence of the NCVS and UCR. New York: Cambridge University Press. Messner, Steven F., Lawrence E. Raffalovich, and Richard McMillan. 2001. Economic deprivation and changes in homicide arrest rates for white and black youths, 1967 – 1998: A national time-series analysis. Criminology 39:591 – 614. National Bureau of Economic Research. 2003. The NBER’s recession dating procedure, 2003. Available at National Bureau of Economic Research. 2010. Business cycle expansions and contractions. Available at Penick, Betty K. E., and Maurice Owens, eds. 1976. Surveying crime. Washington, DC: National Academy Press. Pratt, Travis C., and Francis T. Cullen. 2005. Assessing macro-level predictors and theories of crime: A meta-analysis. In Michael Tonry (ed.), Crime and justice, vol. 32, A review of research. Chicago: University of Chicago Press. Rand, Michael R. 2008. Criminal victimization, 2007. Washington, DC: U.S. Department of Justice. Rand, Michael R., James P. Lynch, and David Cantor. 1997. Criminal victimization, 1973 – 1995. Washington, DC: U.S. Department of Justice. Raphael, Stephen, and Rudolf Winter-Ebmer. 2001. Identifying the effects of unemployment on crime. Journal of Law and Economics 44:259 – 283. Rennison, Callie Marie, and Michael R. Rand. 2007. Introduction to the National Crime Victimization Survey. In James P. Lynch and Lynn A. Addington (eds.), Understanding crime statistics: Revisiting the divergence of the NCVS and UCR. New York: Cambridge University Press. Rosenfeld, Richard. 2009. Crime is the problem: Homicide, acquisitive crime, and economic conditions. Journal of Quantitative Criminology 25:287 – 306. Rosenfeld, Richard, and Robert Fornango. 2007. The impact of economic conditions on robbery and property crime: The role of consumer sentiment. Criminology 45:735 – 769. Sampson, Robert, Stephen Raudenbush, and Felton Earls. 1997. Neighborhoods and violent crime: A multilevel study of collective efficacy. Science 277:918 – 924.

This page intentionally left blank

5 The Nonlinear Effect of Neighborhood Disadvantage on Youth Violence Neighborhood Effects on Youth Violence Xiangming Fang, Richard Rosenfeld, Linda L. Dahlberg, and Curtis S. Florence

Introduction Youth violence is a serious public health problem that affects young people, their families, and communities across the United States. Homicide is the second leading cause of death for young people between the ages of 10 and 24 in the United States (CDC 2009).1 In 2007, 5,764 young people in this age group were victims of homicide, and more than 668,000 cases of violence-related injuries in young people aged 10 to 24 were treated in U.S. emergency departments (CDC 2009). In addition to causing injury and death, youth violence affects communities by increasing health care expenditure, reducing productivity, decreasing property values, and disrupting social services (Mercy et al. 2000). One of the main emphases in contemporary youth violence research is the influence of neighborhood/community disadvantage on fatal and >>





this problem is to control for individual- and family-level factors that may contribute to compositional differences among families and individuals within neighborhoods (e.g., race/ethnicity, family structure, social stratification) or are related to movement into or out of neighborhoods with differing characteristics. Obviously families are not randomly assigned to neighborhoods but make choices about where to live on the basis of factors such as education, economic status, and perceived characteristics of the neighborhood, or they confront barriers to moving to particular neighborhoods, especially on the basis of race/ethnicity or income (Massey et al. 1994; Elliott et al. 1996; South and Crowder 1997; Oakes and Rossi 2003; De Coster, Heimer, and Wittrock 2006). However, Sampson, Morenoff, and Gannon-Rowley (2002) warn that researchers should be careful not to overcontrol for factors that might mediate the link between neighborhood and youth behavior; otherwise, neighborhood effects could be underestimated. Another complexity in examining the impact of community disadvantage on youth violence is the possibility of a nonlinear relationship between measures of disadvantage and violence (i.e., neighborhood effects on violence may not be the same across the entire range of neighborhood conditions). Researchers have pointed out that a high risk of social problems, including violence, appears only in the most disadvantaged communities (Crane 1991; Lauritsen 2003). Winslow and Shaw (2007) examined the impact of neighborhood disadvantage on early child behavior problems in a low-income, urban sample of 281 African American and European American boys and found the impact was nonlinear, reflecting a “threshold effect.” A steep and significant slope was detected only for children living in underclass neighborhoods. Their findings suggest that a threshold of disadvantage has to be surpassed in order for neighborhood disadvantage to have a significant impact on child behavior problems. The results also imply that studies that use continuous measures of neighborhood disadvantage may underestimate neighborhood effects. This study addresses the question of whether the effects of community disadvantage on youth violence are nonlinear. Using a nationally representative school-based sample of U.S. adolescents, we seek to establish whether neighborhood disadvantage is associated with violent behavior by youth, net of compositional and selection effects. Following




the in-home survey, which includes questions on self-reported violent behavior. Of these, 18,924 have a sampling weight. When the sampling weights are incorporated into analyses, the results are considered nationally representative of students between the ages of 11 and 21 years (Chantala and Tabor 1999; Udry 2003). An additional 367 respondents are excluded because they are missing information on outcome variables or main independent variables, resulting in a final study sample of 18,557 participants. Add Health is particularly well suited for this study, due to several desirable features of the data. First, it includes data on individual- and family-level sociodemographic characteristics, census-based measures of neighborhood (tract-level) characteristics, and adolescents’ selfreported involvement in violent behavior. The contextual-level data allow us to link individual violent behavior with a measure of neighborhood disadvantage and to examine the net effect of neighborhood disadvantage on youth violence after adjusting for individual- and familylevel sociodemographic factors. Drawing on a nationally representative sample of adolescents allows us to investigate the impact of neighborhood disadvantage on youth violence across a variety of neighborhood contexts and therefore affords greater generalizability of the study results. The sample of 18,557 adolescents is nested within 2,326 census tracts in 252 counties.

Measures Dependent Variables Our dependent variables, self-reported involvement in violent behaviors, are taken from the Wave I in-home survey. We focus first on a dichotomous (yes/no) measure of general youth violence. Participants are categorized as engaging in violent behavior if they reported committing any of the following offenses during the past 12 months: (1) engaged in a serious physical fight; (2) hurt someone badly enough to need bandages or care from a doctor or nurse; (3) took part in a group fight; (4) used or threatened to use a weapon to get something from someone; (5) pulled a knife or gun on someone; (6) shot or stabbed someone; (7) ever used a weapon in a fight. We then analyze two subcategories of the




Lauritsen 2001), three dimensions were found to describe the sociodemographic structure of U.S. census tracts: socioeconomic disadvantage, immigrant concentration, and residential stability.3 The socioeconomic disadvantage index is a standardized and weighted index combining five variables which load highly on the disadvantage dimension (factor loadings are in parentheses): the proportion of nonelderly with income below the poverty line (0.89), the proportion of family households receiving public assistance income (0.89), the unemployment rate (.83), the proportion of family households that are female headed with children under 18 years of age (.79), and the proportion of the population aged 25+ without a high school diploma or equivalency (0.79).4 Similarly, the residential stability index is a standardized and weighted index combining two variables: the proportion of individuals aged five or older who lived in the same house five years ago and the proportion of owner-occupied homes. The immigration concentration index is a standardized and weighted index combining the proportion of the tract population that is foreign born and the proportion that is of Hispanic origin. The residential stability and immigration concentration indexes are included in our analyses to control for other neighborhood attributes that may contribute to residual confounding bias when examining the impact of neighborhood socioeconomic disadvantage on violent behavior by youth. Other Control Variables To control for compositional effects, we include the following individual- and family-level characteristics in our models: adolescent’s race/ ethnicity, sex (1 = male, 0 = female), grade dummies (to capture cohort effects), whether the adolescent was foreign born (1  =  foreign born, 0 = born in the United States), family structure, primary caregiver education, and the family’s receipt of public assistance (1 = receipt of any public assistance, 0  =  nonreceipt of public assistance). Respondents’ self-identified race/ethnicity was grouped into six categories: White non-Hispanic, Black non-Hispanic, Hispanic, Native American nonHispanic, Asian non-Hispanic, and other. We classified family structure into four categories: living with both biological parents, two-parent families in which one or both parents are stepparents, single-parent




Results Table 5.1 presents descriptive statistics for the variables used in our analyses. After applying the sample weights, the distribution of the study sample by sex and race/ethnicity is very similar to that of the nation as a whole. About 43% of the sample reported participating in at least one of the seven acts of violent behavior during the past 12 months. Although about 41% of the sample reported engaging in fighting, only 10% reported engaging in weapon-related violence. The average age of participants in the study sample was 16 years old, 6.4% of the sample was foreign born, and 54% lived with two biological parents. In addition, there is considerable variation in the primary caregiver’s education and the family’s receipt of public assistance. Logistic Regressions Tables 5.2 and 5.3 present the relationship between neighborhood disadvantage and the dichotomous measure of general violence as blocks of covariates are added to the model sequentially. Model 1 in table 5.2 presents the relationships between the three contextual variables and general violence using the socioeconomic disadvantage index. Model 1 in table 5.3 presents the same relationships using the disadvantage quintile dummies, which allows for inspection of nonlinear effects. The linear relationship between the neighborhood disadvantage index and general violent behavior is positive and statistically significant in the specification shown in Model 1 of table 5.2. The results for the disadvantage dummy variables in Model 1 of table 5.3 indicate no significant relationship between general violence and less disadvantaged neighborhoods (second from the bottom quintile), when contrasted with the least disadvantaged areas (the bottom quintile). The coefficients for the remaining disadvantage dummy variables increase are positive, are significant, and increase in magnitude as the level of disadvantage increases. In both models, residential stability is negatively and significantly associated with general violence, and immigrant concentration has no significant relationship with general violence. Model 2 of tables 5.2 and 5.3 adds region, urbanicity, school type, and school size to the neighborhood context measures. With the exception



ther measure is significantly associated with adolescent reports of general violent behavior net of other individual- and family-level factors. Adding these two variables has little impact on the other results. The effect of neighborhood residential stability persists after the inclusion of the primary caregiver’s age and the selection measure. Tables 5.4 through 5.7 present the results for the dichotomous measures of fighting and weapon-related violence. These tables are formatted identically to tables 5.2 and 5.3, with covariates entered sequentially in blocks. The results for the fighting and weapon-related violence measures are, with few exceptions, similar to those obtained for general violence. Two differences are the nonsignificant effects of residential stability and the middle quintile of neighborhood disadvantage on weapon-related violence when the school, individual, and family covariates are added (see tables 5.6 and 5.7). Nonetheless, although the pattern of disadvantage effects on weapon-related violence differs somewhat from those found for general violence and fighting, we observe a nonlinear relationship between neighborhood disadvantage and weapon-related violence. Spline Regressions Spline regression offers another way to assess the nonlinear relationship between neighborhood disadvantage and violent behavior. Spline regression essentially partitions the relationship into meaningful and discrete linear segments. In other words, spline regression estimates different linear relationships between neighborhood disadvantage and violent behavior for different ranges of neighborhood disadvantage. Based on the logistic regression results presented in tables 5.2 – 5.7, we created three piecewise linear functions (the “splines”) of neighborhood disadvantage using the two “knots” of the 40th percentile and 60th percentile on the disadvantage index. The results of the spline regressions, which are conditioned on all of the individual, family, and neighborhood control variables, are presented in table 5.8. We also graphically depict one of the regression results (general violence) to illustrate the nonlinear relationship (see figure 5.1). For all three of the outcome measures, we find a significant effect (at p 


performed using negative binomial regressions due to the pronounced skew of the index measures. The results are substantively similar to those obtained using the dichotomous outcome measures and clearly indicate the same pattern of nonlinear relationship between neighborhood disadvantage and violent behavior. We report the results of the negative binomial regressions predicting violent behavior using the piecewise linear functions of neighborhood disadvantage in table 5.9. Previous multilevel studies of community disadvantage and adverse individual outcomes, including violent behavior, have included interaction terms to determine whether the effect of disadvantage is conditioned by individual and family factors such as race and family structure. Significant cross-level interactions are often detected (e.g., Lauritsen 2003). We followed suit and tested for interaction effects on violent behavior of neighborhood disadvantage and adolescent race/ ethnicity, family structure, and the family’s receipt of public assistance. When the three interaction terms were entered in our models containing the continuous index measure of disadvantage (Model 4 in tables 5.2, 5.4, and 5.6), we found significant effects for the interaction between disadvantage and race/ethnicity and for the interaction between disadvantage and family structure for all three of the outcome measures but no significant effects for the interaction between disadvantage and receipt of public assistance. The significant results suggest that the relationship between disadvantage and violent behavior by youth is especially pronounced for whites and those living with both biological parents. However, in the piecewise linear spline regression models that account for nonlinear relationships between disadvantage and violent behavior by youth, the interaction terms for disadvantage and race/ethnicity and disadvantage and family structure are no longer significant. This finding suggests that some of the interaction effects previously detected may have resulted from treating the relationship between community disadvantage and adverse individual behaviors as if it were linear.

Discussion Youth violence is a major public health problem in the United States and has been an important focus of research by public health researchers,




analysis suggests that the cross-level interactions between community disadvantage and individual and family characteristics found in prior research may be an artifact of assuming a linear relationship between disadvantage and violence and other adverse behaviors. Our finding of a “step-up” or threshold effect of community disadvantage on violent behavior is consistent with the results of Winslow and Shaw’s (2007) study of the effect of neighborhood disadvantage on boys’ risk for early overt behavior problems and Lauritsen’s (2003) study of adolescent violence. However, the detection of a ceiling effect in the current study differs from their results. For example, Lauritsen found a significant elevation in violence only for the 10% of adolescents living in the most disadvantaged areas. There are a number of possible reasons for the divergence in results, including the use of different methods of analysis and sets of covariates. The most important difference may be sample design. Lauritsen’s study is based on a household sample, whereas the current research uses a school-based sample. Although both are national probability samples, school samples are likely to underrepresent the most violent youth because they have dropped out of school or attend irregularly. Assuming that many of these adolescents reside in highly disadvantaged neighborhoods, their underrepresentation could attenuate the relationship between violence and community disadvantage in the upper tail of the disadvantage distribution and thereby account, at least in part, for the ceiling effect we find. This is an important area for future investigation. The underrepresentation of the most violent youth from schoolbased samples is part of the larger selection problem that has bedeviled cross-sectional multilevel research and even field experiments that have sought to randomly relocate families from high-poverty to low-poverty neighborhoods (see Sampson 2008). Families are not randomly assigned to neighborhoods but rather select neighborhoods based on a number of factors, including their income and what they can afford, the quality of schools, proximity to employment and other amenities, and proximity to family and friends. Although we controlled for several individual and family factors that may be related to the selection of neighborhoods, including a measure of the caregiver’s reasons for choosing the neighborhood in which the family lives, other unmeasured family characteristics such as substance abuse and mental




This is not surprising given that local laws, public transportation issues, zoning restrictions, landlord preferences, and voucher program requirements themselves typically skew available options for families in these disadvantaged areas (Malpezzi 1996; Turner, Popkin, and Cunningham 2000; Turner et al. 2002). In addition, the strength of existing social ties to friends and families should not be underestimated in calculating the distances families will move to take advantage of better housing and opportunities. However, even if families moved to slightly betteroff neighborhoods, the results of our study suggest a “step-up” in violent behavior in areas considered to be in the midrange of disadvantage, as illustrated in figure 5.1. Without knowing the characteristics of these communities, it is conceivable that these are largely transitional communities — either making the transition to more middle-class stability or slowly declining and showing signs of stress. These may be neighborhoods that are not poor or crime ridden enough to receive much in the way of state or federal funding but also are not stable enough to support their own initiatives, as is the case in more prosperous neighborhoods. The transitional nature and potential lack of cohesion in these types of communities also may make them less responsive to community members’ needs and preferences. While our results suggest that moving from the midrange of disadvantage to more stable and prosperous neighborhoods may have a preventive effect on violent behavior, the mechanisms underlying such an effect constitute an important question for future research. In terms of policy and program implications, there are at least three factors to consider. First, as evident in recent examinations of the impact of housing voucher programs, simply providing opportunities to move to better-off neighborhoods does not mean families will necessarily do so or that they will necessarily experience a range of positive outcomes as a result. Effort must also be made to address the constellation and cumulative nature of risk that characterizes impoverished families wherever they might end up and that is ultimately tied to violent behavior among youth. Even within highly disadvantaged neighborhoods, there can be quite a bit of variability in the offending patterns of youth (Gorman-Smith et al. 1998). Youths on a developmental trajectory toward serious violent and delinquent behavior often live in family environments that have a multitude of problems, including




developing and testing strategies aimed at moderating or buffering against risk. It is quite possible to conceive of a mixture of factors that moderate risk for youths and families in communities of disadvantage, either in highly distressed communities or in the midrange of disadvantage. In terms of the broader community context or neighborhoods themselves, it is also quite possible to conceive of a mixture of factors that are protective. Identifying and addressing protective factors as well as the tipping points for individuals, families, and communities may go a long way toward preventing violence and creating healthier, safer, and thriving communities.

Appendix A: Statistical Analysis Logistic regression was used to analyze the relationships between neighborhood disadvantage and violent behavior by youth for the three dichotomous measures of violence: general violence, fighting, and weapon-related violence. To examine the presence of nonlinear relationships between violent behavior and neighborhood disadvantage, we estimated two separate models for each measure of violent behavior. The first model includes the socioeconomic disadvantage index as a continuous measure. For the second model, we divided the sample into quintiles based on respondents’ scores on the socioeconomic disadvantage index and included the quintile dummies as separate independent variables. The same control variables, entered in blocks, are included in each model. We used piecewise linear spline regression models to further examine possible threshold and ceiling effects of neighborhood disadvantage on violent behavior. To assess the robustness of our findings, we repeated our analyses using the three disadvantage variability indexes as the outcome measures. Because the indexes are strongly skewed, discrete variables with variances much greater than the means, negative binomial regressions were used in these analyses. The analyses were performed using Stata SE version 9, which allows for the control of survey design effects of individuals clustered within schools and stratification by geographic region. Poststratification weights were applied to generate nationally representative estimates.




Table 5.2. Logistic Regressions Predicting Youth Violence in General Using the Socioeconomic Disadvantage Index Coefficient (Standard error) Variables

Model 1

Neighborhood context indexes Socioeconomic disadvantage 0.095*** (0.018) Immigrant concentration –0.032 (0.024) Residential stability –0.083*** (0.024) Region (reference category: West) Midwest South Northeast Urbanicity (reference category: urban) Suburban Rural Private school School size (reference category: small) Medium Large Individual characteristics Sex (male) Foreign-born Race/ethnicity (reference category: white) Hispanic Black Native Asian Other Family characteristics Family structure (reference category: two biological parents) Two parents with ≥ 1 nonbiological parent Single parent Other Primary caregiver education (reference category: < high school) High school Some college College graduate Above college Parent receiving public assistance †

Model 2

Model 3†

Model 4†

0.090*** (0.018) –0.040 (0.023) –0.079** (0.025)

0.055** (0.017) –0.048 (0.031) –0.068** (0.023)

0.016 (0.017) –0.043 (0.031) –0.056* (0.022)

–0.081 (0.097) –0.058 (0.089) 0.047 (0.142)

0.004 (0.094) –0.059 (0.091) 0.102 (0.132)

0.018 (0.088) –0.028 (0.091) 0.077 (0.131)

–0.071 (0.086) –0.194 (0.114) –0.302 (0.178)

–0.007 (0.088) –0.067 (0.106) –0.220 (0.161)

0.005 (0.081) –0.083 (0.101) –0.090 (0.149)

–0.058 (0.120) –0.191 (0.123)

–0.015 (0.118) 0.029 (0.124)

0.003 (0.116) 0.057 (0.124)

Regression results for grade dummies are omitted from the table. * p < .05; ** p < .01; *** p < 0.001

0.879*** (0.041) –0.436*** (0.072)

0.900*** (0.042) –0.429*** (0.072)

0.538*** (0.101) 0.562*** (0.074) 0.732*** (0.205) 0.083 (0.109) –0.243 (0.248)

0.449*** (0.102) 0.488*** (0.071) 0.641** (0.205) 0.169 (0.111) –0.299 (0.270)

0.280*** (0.057) 0.315*** (0.047) 0.401* (0.171)

–0.101 (0.073) –0.244** (0.078) –0.545*** (0.092) –0.648*** (0.102) 0.208*** (0.058)




Table 5.4. Logistic Regressions Predicting Fighting Using the Socioeconomic Disadvantage Index Coefficient (Standard error) Variables

Model 1

Neighborhood context indexes Socioeconomic disadvantage 0.091*** (0.018) Immigrant concentration –0.020 (0.025) Residential stability –0.081*** (0.024) Region (reference category: West) Midwest South Northeast Urbanicity (reference category: urban) Suburban Rural Private school School size (reference category: small) Medium Large Individual characteristics Sex (male) Foreign-born Race/ethnicity (reference category: white) Hispanic Black Native Asian Other Family characteristics Family structure (reference category: two biological parents) Two parents with ≥ 1 nonbiological parent Single parent Other Primary caregiver education (reference category: < high school) High school Some college College graduate Above college Parent receiving public assistance †

Model 2

Model 3†

Model 4†

0.085*** (0.018) –0.023 (0.024) –0.080** (0.024)

0.057** (0.018) –0.032 (0.031) –0.068** (0.023)

0.019 (0.017) –0.028 (0.030) –0.058** (0.022)

–0.060 (0.098) –0.071 (0.090) 0.059 (0.142)

0.018 (0.096) –0.070 (0.092) 0.111 (0.132)

0.032 (0.091) –0.038 (0.093) 0.087 (0.132)

–0.061 (0.084) –0.178 (0.112) –0.314 (0.180)

0.006 (0.086) –0.052 (0.104) –0.218 (0.168)

0.018 (0.081) –0.069 (0.101) –0.093 (0.157)

–0.066 (0.123) –0.212 (0.126)

–0.013 (0.121) 0.028 (0.131)

0.005 (0.120) 0.055 (0.131)

Regression results for grade dummies are omitted from the table. * p < .05; ** p < .01; *** p < 0.001

0.821*** (0.040) –0.407*** (0.072)

0.839*** (0.041) –0.397*** (0.072)

0.498*** (0.100) 0.484*** (0.073) 0.647** (0.221) 0.072 (0.113) –0.348 (0.259)

0.413*** (0.102) 0.415*** (0.069) 0.558* (0.217) 0.159 (0.114) –0.397 (0.278)

0.257*** (0.058) 0.276*** (0.046) 0.363* (0.178)

–0.088 (0.076) –0.221** (0.079) –0.528*** (0.094) –0.629*** (0.100) 0.211*** (0.058)




Table 5.6. Logistic Regressions Predicting Weapon-Related Youth violence Using the Socioeconomic Disadvantage Index Coefficient (Standard error) Variables

Model 1

Neighborhood context indexes Socioeconomic disadvantage 0.114*** (0.018) Immigrant concentration –0.063 (0.038) Residential stability –0.094* (0.038) Region (reference category: West) Midwest South Northeast Urbanicity (reference category: urban) Suburban Rural Private school School size (reference category: small) Medium Large Individual characteristics Sex (male) Foreign-born Race/ethnicity (reference category: white) Hispanic Black Native Asian Other Family characteristics Family structure (reference category: two biological parents) Two parents with ≥ 1 nonbiological parent Single parent Other Primary caregiver education (reference category: < high school) High school Some college College graduate Above college Parent receiving public assistance †

Model 2 0.121*** (0.019) –0.120** (0.041) –0.065 (0.041)

Model 3†

Model 4†

0.054** (0.019) –0.107 (0.054) –0.059 (0.039)

0.023 (0.019) –0.094 (0.053) –0.036 (0.038)

–0.119 (0.170) 0.040 (0.158) 0.174 (0.178)

0.012 (0.148) 0.051 (0.135) 0.261 (0.150)

0.039 (0.138) 0.104 (0.129) 0.260 (0.146)

–0.191 (0.110) –0.322 (0.171) –0.073 (0.187)

–0.157 (0.101) –0.239 (0.149) –0.093 (0.149)

–0.143 (0.096) –0.232 (0.145) 0.018 (0.141)

0.104 (0.135) 0.167 (0.143)

0.054 (0.120) 0.129 (0.134)

0.063 (0.118) 0.138 (0.132)

1.126*** (0.082) –0.727*** (0.156)

1.142*** (0.083) –0.715*** (0.153)

0.667*** (0.134) 0.784*** (0.094) 1.186*** (0.284) 0.225 (0.177) 0.438 (0.344)

0.610*** (0.134) 0.639*** (0.094) 1.087*** (0.292) 0.311 (0.177) 0.377 (0.344)

Regression results for grade dummies are omitted from the table. * p < .05; ** p < .01; *** p < 0.001

0.403*** (0.083) 0.521*** (0.083) 0.985*** (0.206)

0.059 (0.113) –0.088 (0.111) –0.136 (0.156) –0.350 (0.180) 0.203* (0.078)




Table 5.8. Effect of Neighborhood Disadvantage on Youth Violence (Measured as Dichotomous Variables): Results of Piecewise Linear Spline Regressions Coefficient (Standard error) Variables Splines of neighborhood disadvantage Least to less neighborhood disadvantage† Middle neighborhood disadvantage† More to most neighborhood disadvantage†

YV in general


Weapon-related YV

0.037 (0.078) 0.021 (0.083) 0.056 (0.130) 0.421** (0.137) 0.458** (0.145) 0.550** (0.202) –0.036 (0.021) –0.033 (0.022) –0.034 (0.028)

Note: YV = youth violence † Estimates adjusted for individual, family, and neighborhood-level covariates. * p < .05; ** p < .01; *** p < 0.001

Table 5.9. Effect of Neighborhood Disadvantage on Youth Violence (Measured as Variability Indexes of Youth Violence): Results of Piecewise Linear Spline Regressions Variables Splines of neighborhood disadvantage Least to less neighborhood disadvantage† Middle neighborhood disadvantage† More to most neighborhood disadvantage†

Coefficient (Standard error) YV in general Fighting Weapon-related YV 0.043 (0.065) 0.051 (0.058) 0.015 (0.153) 0.276** (0.096) 0.214* (0.088) 0.548* (0.220) –0.010 (0.014) –0.006 (0.013) –0.022 (0.029)

Note: YV = youth violence † Estimates adjusted for individual, family, and neighborhood-level covariates. * p < .05; ** p < .01; *** p < 0.001

Notes 1. Disclaimer: the authors report no conflicts of interest. The findings and conclusions in this chapter are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention. 2. World Health Organization, “Health Topics: Violence,” 2012, topics/violence/en/. We do not address violence “against oneself,” or suicide, in this study. 3. Similar to Lauritsen’s (2001) findings, age composition (the percentage of the population less than age 18), which loads on the socioeconomic disadvantage index in Sampson et al.’s (1997) study, was found to be a single-item factor. However, this variable was not significantly associated with violent behavior, regardless of the measure of violent behavior and the specific regression model used, and is not included in the analyses shown. 4. The factor loading for the percentage of black residents is 0.55 and is not included in the disadvantage index reported here. We conducted analyses with percentage black included in the disadvantage index and obtained results nearly identical




poverty, vol. 1, Contexts and consequences for children, 62 – 78. New York: Russell Sage Foundation. Elliott, Delbert S., William Julius Wilson, David Huizinga, Robert J. Sampson, Amanda Elliott, and Bruce Rankin. 1996. The effects of neighborhood disadvantage on adolescent development. Journal of Research in Crime and Delinquency 33:389 – 426. Feshbach, Seymour. 1980. Child abuse and the dynamics of human aggression and violence. In George Gerbner, Catherine J. Ross, and Edward Zigler (eds.), Child abuse: An agenda for action. New York: Oxford University Press. Gephart, Martha A. 1997. Neighborhoods and communities as contexts for development. In Jeanne Brooks-Gunn, Greg J. Duncan, and Lawrence Aber (eds.), Neighborhood poverty, vol. 1, Context and consequences for children, 1 – 43. New York: Russell Sage Foundation. Gorman-Smith, Deborah, Patrick H. Tolan, Rolf Loeber, and David B. Henry. 1998. Relation of family problems to patterns of delinquent involvement among urban youth. Journal of Abnormal Child Psychology 26 (5): 319 – 333. Haynie, Dana L., Eric Silver, and Brent Teasdale. 2006. Neighborhood characteristics, peer influence, and adolescent violence. Journal of Quantitative Criminology 22 (2): 147 – 169. Hipp, John R. 2007. Block, tract, and levels of aggregation: Neighborhood structure and crime and disorder as a case in point. American Sociological Review 72 (5): 659 – 680. Katz, Lawrence F., Jeffrey R. Kling, and Jeffrey B. Liebman. 2001. Moving to Opportunity in Boston: Early results of a randomized mobility experiment. Quarterly Journal of Economics 116:607 – 654. Kornhauser, Ruth. 1978. Social sources of delinquency. Chicago: University of Chicago Press. Krivo, Lauren J., and Ruth D. Peterson. 1996. Extremely disadvantaged neighborhoods and urban crime. Social Forces 75:619 – 650. Lauritsen, Janet L. 2001. The social ecology of violent victimization: Individual and contextual effects in the NCVS. Journal of Quantitative Criminology 17 (1): 3 – 32. Lauritsen, Janet L. 2003. How families and communities influence youth victimization. NCJ 201629. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention. Ludwig, Jens, Greg J. Duncan, and Paul Hirschfield. 2001. Urban poverty and juvenile crime: Evidence from a randomized housing-mobility experiment. Quarterly Journal of Economics 116:655 – 679. Ludwig, Jens, Jeffrey B. Liebman, Jeffrey R. Kling, Greg J. Duncan, Lawrence F. Katz, Ronald C. Kessler, and Lisa Sanbonmatsu. 2008. What can we learn about neighborhood effects from the Moving to Opportunity experiment? American Journal of Sociology 114 (1): 14 – 88. Malpezzi, Stephen. 1996. Housing prices, externalities, and regulation in U.S. metropolitan areas. Journal of Housing Research 7 (2): 209 – 241.




Turner, Margery Austin, Susan J. Popkin, and Mary K. Cunningham. 2000. Section 8 mobility and neighborhood health. Washington, DC: Urban Institute. Turner, Margery Austin, Stephen Ross, George Galster, and John Yinger. 2002. Discrimination in metropolitan housing markets: Results from phase I of HDS2000. Washington, DC: U.S. Department of Housing and Urban Development. Udry, J. Richard. 2003. The National Longitudinal Study of Adolescent Health (Add Health), Waves I & II, 1994 – 1996; Wave III, 2001 – 2002 [MRDF]. Carolina Population Center, University of North Carolina at Chapel Hill. U.S. Department of Health and Human Services. 2001. Youth violence: A report of the Surgeon General. Rockville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Injury Prevention and Control; Substance Abuse and Mental Health Services Administration, Center for Mental Health Services; and National Institutes of Health, National Institute of Mental Health. Winslow, Emily B., and Daniel S. Shaw. 2007. Impact of neighborhood disadvantage on overt behavior problems during early childhood. Aggressive Behavior 33:207 – 219.

6 Aggravated Inequality Neighborhood Economics, Schools, and Juvenile Delinquency Robert D. Crutchfield and Tim Wadsworth

One of the challenges of the scholarship connecting macroeconomic conditions to youth violence has been identifying the mechanisms that underlie the process by which individuals are influenced by the structural conditions in their communities.1 While there has been much focus on labor markets and the impact of unemployment as well as employment in secondary sectors, a key theoretical connection has been missing. The people who have been directly influenced by unemployment and macro shifts in types of available employment (e.g., labor market restructuring resulting from deindustrialization) have been adults, while much of the violent crime is committed by adolescents or young adults who have not yet completed the transition to full-time 152



work. This chapter addresses this issue by examining how concentrated disadvantage at the neighborhood level can both precede and condition the relationship between adolescents’ attachment to school and their participation in delinquent behavior. Here we focus on neighborhood disadvantage, a consequence of long-term macroeconomic patterns and one of three means by which the labor market may affect youth behavior. The other two, the evidence of which will be briefly reviewed later, are juvenile employment and the work experience of parents. It is not hard to imagine that economic disadvantage may cause adults to engage in criminal behavior. Traditional sociological and criminological theories offer a variety of mechanisms connecting disadvantage and crime, and recent empirical work has debunked some and affirmed other hypotheses drawn from these theories (Fagan and Freeman, 1999). While popular conceptions of the role of unemployment in shaping criminal behavior often suggest that unemployed individuals engage in crime to provide or supplement their income, we know that the association between work and crime is neither simple, consistent, nor strong (Freeman, 1983; Cantor and Land, 1985; Parker and Horwitz, 1986; Box, 1987; Chiricos, 1987). As a result, researchers have sought more complete explanations of how employment is linked to adult criminality at both the individual and aggregate levels (Fagan and Freeman, 1999; Crutchfield and Pitchford, 1997; Uggen, 1999; Wadsworth, 2000, 2004; Gould, Weinberg, and Mustard, 2002; Grogger, 2006). These explanations focus on more nuanced relationships between individuals, communities, and labor markets. Children’s lives and their relationships to the labor market and the economy are even less straightforward. Their relationship to the economy is primarily through their parents, and their status is tied either to family socioeconomic status or to the unique hierarchy that is established among adolescents. Unemployment or low-quality employment, both of which are often viewed as more proximate predictors of crime for adults, mean something very different for adolescents. We suggest, however, that economic and labor market factors likely do influence children’s criminality but that their effects are more indirect and contextual in nature. For example, instead of a direct effect of family poverty on children’s crime, we believe that the influence may operate




Warren et al.’s argument is the degree to which an adolescent’s employment interfered with his or her educational achievement and aspirations. They reported that juveniles who were already not attached to school and those who worked longer hours were more likely to be involved in delinquency (the selection argument made by others) but that those who had good grades and worked more modest hours were actually less likely to become involved in criminal behavior. Warren et al.’s findings highlight an important approach to understanding the relationship between labor markets and delinquency: the role of education as a mediator between youth employment and delinquency, as well as between labor market conditions and delinquency. Indeed, Staff and Uggen (2003) found that work that “supports” school (employment cultures which stress or encourage academic success) leads to less delinquency and drug use, but jobs that are more “adult like” have the opposite effects.

Adult Employment and Juvenile Delinquency As early as 1989, Sullivan reported in Getting Paid that the employment that was available to adults in communities was linked to the amount and type of delinquency in which young people there became engaged. Crutchfield, Rankin, and Pitchford (1993) found that marginal parental employment and local joblessness demonstrated small but significant indirect effects on delinquency, operating through the academic performance of children. Wadsworth (2000) found that this link was related in part to children’s diminished academic expectations, their lower feelings of efficacy, and less adult supervision when the parent was employed in a marginal setting, and Bellair and Roscigno (2000) found that parents’ low-wage jobs and unemployment led to children’s problem behavior, which they attributed to resulting problems in families. Bellair, Roscigno, and McNulty (2003) reported that parents’ labor market experience indirectly influenced their children’s delinquency through their school performance. These findings have been interpreted to mean that children invest less in education when the evidence in front of them, the work experience of parents, suggests that “playing by the rules” — or, to use Merton’s (1949) term, “conformity” — offers little hope for advancement or




This process had a profound effect on both individuals and communities. Scholars have drawn primarily on social control theory to explore the effect this shift had on individuals, arguing that unemployment, as well as unstable or unrewarding work, diminished the “stakes in conformity” that deterred problematic and criminal behavior (Toby, 1957; Hirschi, 1969; Crutchfield, 1989; Sampson and Laub, 1993, Wadsworth, 2006). This negative consequence intensified when the destabilization of labor markets affected whole communities. When the number of people with no, or less rewarding, jobs reaches a critical level, a “situation of company” is created in which large numbers of occupationally unattached individuals, unconstrained by the requirements of a steady job, congregate in public spaces such as in bars and on street corners. Such situations, it has been suggested (Crutchfield, 1989), are often conducive to criminal behavior. Anderson (1999) has also suggested that such social processes can give rise to oppositional cultures which are more tolerant of, and at times encourage, deviant and criminal behavior. These connections between labor market segmentation and criminal participation have been empirically demonstrated at both the individual level (Crutchfield and Pitchford, 1997; Sampson and Laub, 1993; Uggen 1999; Wadsworth, 2006) and the aggregate level (Crutchfield, 1989; Krivo and Peterson, 2004; Wadsworth, 2004). Determining that the quality of employment, and its location within the larger economic structure, influences the criminality of adults is an important contribution to the criminological literature. However, it leaves out a central piece of the puzzle. Many of the people who are directly affected by economic and labor market shifts are adults, while much urban crime is committed by juveniles. Most juveniles have not yet fully entered the labor market, and frequently the only jobs that are available to most of them are low-quality, secondary-sector jobs (the same types of jobs that may be criminogenic for adults). So, while there is evidence that unemployment, as well as low-quality and secondarysector work, is positively related to criminal involvement for young adults (Crutchfield and Pitchford, 1997; Krivo and Peterson, 2004; Wadworth, 2006), things are less clear for kids. Despite the growing literature on adolescent employment that was reviewed earlier, we know less about how labor market patterns influence juvenile delinquency than we do about how they influence adult criminality.




1995; Crutchfield and Pitchford, 1997), county data were used to test the linkage between macro labor market forces, adult work experience, and crime. However, given the thesis being tested here, it would be more appropriate to use neighborhood rather than county data to examine macro effects. Census tract data for NLSY respondents are now available, albeit under very controlled circumstances. We use these data for our analyses. The Bureau of Labor Statistics (BLS) and the Center for Human Resource Research (CHRR) do not release the geocoded census tract information, even with the confidentiality assurances that they require with the county data. Their justifiable concern is that with the geocoded data, it would be possible to identify respondents. Clearly this concern is magnified when the aggregate data appended to individual respondents is at the tract level. Consequently these data could only be used by staff, on site at the CHRR in Columbus, Ohio, when these analyses were conducted. We recognize that even census tract data are not ideal for our purposes because their boundaries are somewhat arbitrarily drawn and do not necessarily coincide with communities or neighborhoods as they are understood by local residents. However, while not perfect, census tracts are approximately the right size, the Bureau of the Census has endeavored to draw them so that they are as consistent to neighborhoods as possible, and these tracts provide the opportunity to study the multilevel influence of the labor market on crime (Elliot, 1999). By using census tracts as proxies for neighborhoods, we can study how neighborhood context influences the relationship between adolescent school achievement/attachment and criminal behavior.

Analytic Strategy The analytic strategy was driven by our interest in linking individual or micro-level processes with those at the aggregate or macro level. The present research improves on earlier work by using census tract, rather than county-level, data. This is not to say that processes at the county level, such as the distribution of types of industries that characterize local labor markets, are not important; but counties can be quite heterogeneous, and such macroeconomic patterns influence some neighborhoods quite differently than others.




regression instead (multilevel models cannot be run without access to the original data). However, aside from data limitations, there is another reason why using OLS regression may be more appropriate. In the Children of the NLSY data set, there are only one or two respondents per census tract.2 As such, HLM would not be appropriate. Since these nonclustered data do not violate the assumption of data independence, using OLS is not a problem. That said, since a small number of tracts will have two respondents, we will take care to interpret results conservatively.

Variables Our dependent variable of interest is an index of juvenile delinquent behavior. The following behaviors by respondents were used to create the index: skipped school without permission, damaged property of others on purpose, got into a fight at work or school, took something without paying for it, took something worth under $50 not from a store, used force to get money from someone, hit or seriously threatened someone, attacked someone to hurt or kill, tried to con someone, took a vehicle without permission of the owner, broke into a building or vehicle to steal or look, knowingly held or sold stolen goods, helped in a gambling operation, hurt someone enough to need a doctor, lied to parents about something important, and misbehaved so that parent had to come to school. For all of these items, respondents were asked if they had engaged in this behavior (yes or no) in the past year. The mean of respondents’ answers to these questions was used as their delinquency score if values were available for at least eight of the questions. Individual-level independent variables include gender, age, race/ethnicity (dummy variables indicating whether the respondent is black or Hispanic, with white as the reference), family composition (presence of father or stepfather), family’s poverty status, mother’s educational achievement, academic history (including grades, a scale measuring school attachment, and parental involvement in the school), respondent’s employment status (whether the juvenile was employed at the time of the survey), and mother’s employment status. Census tract data including racial composition (percentage black and percentage Hispanic), percentage marginal work force (a combination







employment and crime can be conservatively interpreted to mean that we might want to worry less about the effects of a mother’s work and focus instead on the amount or quality of parental involvement (as well as the family’s poverty status, as will be discussed later). The central predictors of delinquency for the juveniles in this sample are, not surprisingly, family poverty and school attachment/performance. Both respondents’ attachment to school (b  =  – .225, p 





affect the identity development process for adolescents who must negotiate their self-construction within that context.

Street Markets and Youth Violence: General Issues “Street markets” are by definition social spaces, typically within an urban context, centered around the distribution and public sale of illegal goods — drugs, stolen merchandise, and other goods. The common depiction is that they flourish within marginalized communities and social groups excluded from fruitful participation in the mainstream economy. Because such markets exist outside the formal social and regulatory systems of control, they are a site for the production of alternative (“informal”) social structures, sociocultural codes, and practices. Moreover, because the mechanisms of social process and control are often neither institutionalized nor durable (with some exceptions), and the potential rewards lucrative, they are also situations of high risk (with respect to violence) and volatility. The literature on street markets includes a long legacy of “street ethnographies,” including more recent work specifically relevant to youth violence, as well as analyses of gun use and availability, neighborhood and population characteristics where street markets exist, gang involvement, drug markets, and “street codes” or behavioral norms arising out of street market contexts. Much of this literature is highly relevant with respect to the connection between macroeconomic factors and youth violence because (a) a range of macroeconomic factors contributes to social exclusion and marginalization which undergirds the creation of street markets; and (b) it is adolescent/young males who are most directly involved in the violence associated with these contexts. Generally, most research and literature on this issue focuses on the proximal connections between street markets and youth violence as opposed to the role of macroeconomic factors in creating/shaping street markets in the first place. The proximal links established in the literature between street markets and youth violence can be summarized as follows: • Street markets focused on drug sales are the most prone to violence due to the concentrated presence and use of guns.




Fig. 7.1. Street markets, youth identity development, and violence — a model

Street Markets, Youth Violence, and Identity: Current Literature Research concerning the impact of street market contexts on youth identities connected to violence has generally followed the William Julius Wilson thesis (1987) down to more micro levels, including classic work by Elijah Anderson on “codes of the street” and their origin (1990, 1992, 1999; Stewart & Simons 2006). Essentially, the argument is as follows: where economic opportunities are very limited and there is a historical pattern of disconnection from mainstream economic activity, drug selling and other aspects of the “street economy” become the dominant playing field for achievement, material gain, and status (see also Bourgois 1989, 1996b; Fagan 1992; Fagan and Wilkinson 1998; Edberg 1992, 1998, 2007) and thus have a strong role in the development and perpetuation of norms and attitudes about violence. Wilkinson’s and Katz’s rich data from interviews with violent offenders explore the ways in which violence — and gun violence in particular — becomes such an important tool for negotiating personal status (Wilkinson 2004;




ture, a social context shaped by alienation and discrimination has contributed to a situation in which violence becomes an important means of constructing racial/ethnic and gender identities. Moreover, while the street-code literature has focused on young males, Jones (2008) conducted an ethnographic study with inner-city girls, finding that they also come to organize their social world around three key aspects of the code: reputation, respect, and retaliation (also see Miller 2001; Miller & Decker 2001). Cross-culturally, research in Oslo, Norway (Sandberg 2009a, 2009b) highlights two discursive modes of self-presentation among minority drug dealers. One mode focuses on personal narratives of oppression, unemployment, racism, and other difficulties — utilized for personal justification of drug dealing and violence as well as during interaction with welfare agencies. The other, called “gangster discourse,” presents the narrator as “thick-skinned, smart and sexually alluring” (Sandberg 2009a, 532) and is the dominant discourse on the street, used to gain respect. In other related work, Sandberg and colleagues have more generally examined the acquisition of symbolic capital among ethnic-minority/immigrant youth in urban Oslo, where such capital is tied to violent street culture (Sandberg & Peterson 2009). One more aspect of “identity” that deserves exploration is the impact of high poverty and chronic violence on adolescent constructions of risk. This is a phenomenon that occurs with respect to other health risks (see Edberg 2004a, 2007). Essentially, beliefs about a limited future in such high-risk settings change the interpretation of risk, such that the benefits of immediate or short-term social recognition and the resources that may accompany that recognition outweigh concerns over incarceration, injury, or death. In support, DuRant et al. (1995) examined social and psychological factors associated with weapon carrying by African American adolescents in a community with extensive poverty and violence, finding that, among other factors, weapon carrying was significantly associated with a self-appraised probability of being alive at age 25. There is also a body of work that addresses person-environment dynamics contributing to the kinds of resources youth have and choices youth are able to make in situations of high risk. Self-concept models have attempted to understand person-environment dynamics behind clusterings of risk behavior and risk factors, focusing on the construct




Theokas et al. 2005; Lerner et al. 2005) address the presence of protective factors in the youth environment as part of a person-context, developmental process.

The Link to Identity Of key importance for this chapter, the compelling nature of street market conditions in concentrated poverty communities, as a social ecology, exerts a particular effect at the time adolescents are in a key phase of development — thus compounding the effect by linking the “imperatives of the street” to the development of personal identity. Street markets themselves become a unique, integrative social formation that generates identities connected to violence-supportive rationales, motivations, and behavior as an integrative process separate from the impact of exposure to identified risk or protective factors that may have preceded involvement in such contexts — a quality that will be referred to herein as generative.2 There is a contrast between this approach and the risk/ protective-factors models, as well as a complementary aspect. Focusing on a context-based generative process implies less scrutiny of the specific micro-paths that lead to youth involvement in a street market setting and more on what happens when they are there — because of what street markets entail. Youth violence connected to street markets thus reflects a context-based weltaunschuung that may originate from exposure to risk factors yet is generative of behavior on its own terms. A few examples may help illustrate the approach. In an ethnographic study recently conducted in the U.S.-Mexico border region — primarily in high-poverty urban areas such as Ciudad Juarez — investigating perceptions of the “narcotrafficker image” as portrayed in popular media (Edberg 2001, 2004a, 2004b), many youth in focus groups and interviews expressed a desire to have hero songs (in Mexican and border culture, these are called corridos) written and sung about them, like the songs sung about the exploits of popular narcotraffickers. The fact that the exploits described in these narcocorridos included violence and other “risk behaviors” did not deter that desire. By contrast, the violence and other risks foregrounded in these narratives were actually the key to understanding a social context in which the narcotraffickerprotagonists were positioned as individuals who stood out, who possessed




to specific risk factors as defined in the literature becomes less meaningful either in terms of behavioral influence or as an intervention target. In a high-risk context, violence may have many positive attributes. It may be instrumentally useful, worth the risks involved, customarily practiced, and socially valued (even expected) as a means of gaining status, reputation, household support, material benefits, or partners (Bourgois 1989, 1996a; Karandinos 2010). More important, engaging in violent behavior is an integral part of the setting — not an outlier behavior. Thus, the characteristics of the context create or generate ongoing motivations and rationales for engaging in violence. This is qualitatively different from a risk-factor, “exposure-push” process. Using the analogous term from economics, there is a “demand-pull” mechanism involved that comes from the aggregate impact of the social context and its motivational structure. Furthermore, however tempting it may seem, “context” cannot properly be construed as a risk factor itself in the manner that risk factors are typically treated in the literature. Context is not a discrete factor. It is an aggregate phenomenon, whose social force derives from the dynamic interaction of internal characteristics and their articulation with larger, external, socioeconomic and cultural patterns. While, as noted, there are context-based internal processes that motivate and sustain violent behavior, these processes also interact — arguably primarily — with broader phenomena that transcend a given high-risk context. Violent behavior itself, for example, is not new, nor is it always proscribed. Sometimes it is broadly admired — and certainly so in the American cultural idiom. The interplay between a behavior that is socially valued in society as a whole — at least in certain forms — and its interpolation within high-risk contexts where there are fewer behavioral options to place it in perspective very likely amplifies the “demand-pull” effect.

Linking Generative Context, Adolescent Identity Development, and Violence Because the discussion concerns youth violence, the relationship between high-risk contexts and behavior must be considered in terms of one of the key developmental processes for youth and adolescents — the construction of an identity (or of a self ). The generative role of these




sented. In addition, the individual must then validate the identity by “follow-up actions” that are in keeping with one’s purported identity. Regarding youth violence, this dynamic is well described by Wilkinson (2004), Fagan & Wilkinson (1998), Anderson (1990, 1999), and Sandberg (2009a, 2009b) in terms of performance of self within the code of the street and within a set of opposing and value-laden identities — for example, the highly valued “tough” or “wild” identity versus the devalued “punk” and “herb” identities (Wilkinson 2004). There is a consistency across cultures with respect to such identity categories. The street identities just referred to are remarkably consonant with the requirements for a personal reputation within the “reputation versus respectability” identity continuum first identified by P.  J. Wilson (1973) in the Caribbean and amplified by Whitehead (Whitehead 1992; Whitehead, Peterson, & Kaljee 1994) as a framework for understanding male-gendered risk behavior in a context of historical and structural constraints. Thus, extensive involvement in violence and other risk behavior by adolescents is likely to be consistent with the construction and performance of a “risky self ” (or “delinquent self ” or “violent self,” etc.) that is understood as such both by the individual taking on this identity and by others within the high-risk social context in which he or she enacts it. Performance, then, entails encoding one’s actions such that they are consonant with that identity. Encoding occurs in the actions taken — the violence itself and specific symbolic aspects of its presentation. For example, turning the gun sideways is by now a well-known symbolic gesture that has become a part of what could be called “gangsta body language.” The symbolic aspects of violence are also encoded in narrative representations about violence and other risk behavior that become part of the performance — having stories to tell that will have meaning and resonance with other adolescents in terms of creating a socially valued identity and through which feedback can be obtained. Importantly with respect to the impact of context, one cannot invent out of whole cloth the material from which an acceptable identity can be constructed and the types of actions that need to be included. This material will come largely from public models, representations and narratives (public discourse) that exist within and across sociocultural groups and contexts, including narratives that have resonance with




these narco-tales reference at least one meaning-laden framework for organizing actions and provide a kind of narrative road map for becoming a “notable person.”6 Almost all the youth interviewed at a youth prison in Juarez, Mexico, for example, wanted to have a corrido written about them. Thus, the narratives were motivators and meaningful structuring agents for action. As narrative structures, corridos and narcocorridos link behavior and meaning in a kind of semiotics of self. While no one in any interview directly attributed any specific actions or activities that they were or had been involved in, or specific actions of others, to narcocorridos per se, they clearly provided sociocultural material — an objectified self through which an individual could personify the values represented. This is translated into clothes, body language, a physical presentation of self, an attitude toward death and risk, a preoccupation with projecting power and importance, and other ways in which the identity is articulated with daily life in the high-poverty, and high-risk, border setting. It is important to emphasize here that understanding violence and risk behavior from this perspective offers a more holistic picture of the integration of attitudes and behaviors within a social setting, rather than treating those attitudes as distinct factors (e.g., “social norms”), subject to change outside of the context that gives them force.

Implications for Policy and Practice If the function presented here of generative, structural context in framing, motivating, and sustaining youth violence in an interactive process with identity formation is accurate, then one key implication is that the many risk/protective-factor and related prevention approaches are likely to be most useful at an earlier stage, in preventing youth from involvement, or entrenchment, in a high-risk social context — although, as noted, none of these approaches addresses the structural context that underlies the co-occurrence of risk factors. Once youth are involved, however, the context and its imperatives may be more useful as defining points for intervention. Also, within a risk/protective-factors framework, even program models that address community risk factors such as gun and drug availability or peer risk factors such as social norms supporting violence have to be understood as important but incomplete




status. A modified, exploratory use of the “possible selves” approach may be useful for this purpose, tied to the attributes elicited from the first research strategy. • Interventions could employ a range of strategies to introduce nonviolence-related roles8 and behaviors (to the “inventory” of violencerelated behaviors) that satisfy the generative schema as identified. Socialcognitive modeling and positive youth development (PYD) community participation approaches are two such strategies. Another, less tested approach may be to use workshops such as those implemented by the Alternatives to Violence Project (AVP),9 originally a prison-based intervention model that seeks to develop personal mastery and responsibility without violence. • Interventions could also identify the resources and contextual factors necessary to introduce non-violence-related roles/identities/behaviors as realistic possibilities for youth in high-violence communities. This may involve collaboration with schools, businesses, or nonprofit and other funders to support tailored positive youth development and particularly job-related options. The idea is to dilute the potency of violence as a socially validated component of identity by adding, over time, to the social fabric meaningful social roles and employment/economic possibilities that produce visible evidence of personal efficacy and gain and that share some aspects of the qualities identified as worthy of emulation. • Interventions could identify modes of representation and channels of communication through which alternative identities (not dependent on violence) and behaviors identified from the formative research could be disseminated. From that step, social marketing and diffusion approaches could be employed to disseminate messages about these “status-worthy” alternative behaviors and identities.

A distinguishing characteristic of this kind of program approach is that the potential social roles/identities offered are context driven, structured around the complex of behavioral motivations characteristic of specific social contexts, based on an understanding that these motivations are now a product of the context itself, not just the designated risk factors that may have contributed to its existence. Thus, instead of implementing a preset school-based curriculum that teaches conflictresolution skills based on a risk-factor assessment that identifies these




Notes 1. National Center for Injury Prevention and Control (NCIPC), Division of Violence Prevention. 2. To engage with new developments in the social sciences (Bourdieu & Wacquant 1992; Bourdieu 2000; Foucault 1978), street markets contribute to the creation of a “habitus” (Bourdieu 1977) or a “subjectivity” (Foucault 1978), which refer to a conscious and preconscious “way of being in the world” that is best understood as being framed by one’s larger historical era and social formation, including forms of “governability” — e.g., policies and institutions, nongovernment initiatives, structural and political forces (including markets), cultural ideological forces (including religion), and discourses that constitute a contingent field. 3. Sampson and colleagues (Sampson, Morenoff, & Raudenbush 2005; Sampson 2003) and findings from the Project on Human Development in Chicago Neighborhoods have been instrumental in highlighting neighborhood-level and social contextual mediating factors in violence and health risk and in calling for more research to understand these processes. 4. Anthropologist Dorothy Holland (Holland et al. 1998) uses the term “figured world” to refer to the constructed, imagined, meaning-laden, and relational aspects of particular sociocultural spaces — the idea of intersubjectivity. 5. The formative and cognitive role of narrative is well recognized in psychology, child development, and anthropological theory about the self — e.g., Budwig & Wiley 1995; Sperry & Sperry 1995; Mintz 1995; Miller et al. 1992; Ochs & Capp 1996; Bruner 1986). As Ochs and Capp (1996) have noted, “as narratives are apprehended, they give rise to the selves that apprehend them” (p. 22). Indeed the psychologist Jerome Bruner (1986) has called narrative one of the two fundamental modes of cognitive thinking. 6. According to Ochs and Capps (1996), “Narratives situate narrators, protagonists, and listeners/readers at the nexus of morally organized past, present, and possible experiences” (p. 22, also citing Guignon 1993; Heidegger 1962). 7. This has been described elsewhere as the “generative schema” associated with violence or other risk behavior (Edberg 2007). 8. In other words, “possible selves.” 9. See the AVP website:

References Anderson, E. (1990). Streetwise: Race, class, and change in an urban community. Chicago: University of Chicago Press. Anderson, E. (1992). The story of John Turner. In A. V. Harrell & G. E. Peterson (Eds.), Drugs, crime, and social isolation (pp. 147 – 180). Washington, DC: Urban Institute Press.




hepatitis C among homeless drug-injecting youth in San Francisco. Human Organization, 63(3): 253 – 264. Bruner, J. (1986). Actual minds, possible worlds. Cambridge: Harvard University Press. Budwig, N., & Wiley, A. (1995). What language reveals about children’s categories of personhood. In L. Sperry & P. Smiley (Eds.), New directions for child development: Vol. 69. Exploring young children’s concepts of self and other through conversation (pp. 21 – 32). San Francisco, CA: Jossey-Bass. Cannon, E., & Wilkinson, D. (2007, November 13). Crime as self-help among violent adolescents in disadvantaged neighborhoods: a mode of preserving reputation. Paper presented at the annual meeting of the American Society of Criminology, Atlanta, GA. Carrithers, M., Collins, S., & Lukes, S. (Eds.). (1985). The category of the person: Anthropology, philosophy, history. Cambridge: Cambridge University Press. Catalano, R. F., & Hawkins, J. D. (1995). Risk-focused prevention: Using the social development strategy. Seattle: Developmental Research and Programs. Chong, Vincent, Khatharya Um, Monica Hahn, David Pheng, Clifford Yee, and Colette Auerswald. 2009. Toward an intersectional understanding of violence and resilience: An exploratory study of young Southeast Asian men in Alameda and Contra Costa County, California. Aggression and Violent Behavior, 14(6), 433 – 506. Delgado, M. (2002). New frontiers for youth development in the 21st century: Revitalizing and broadening youth development. New York: Columbia University Press. Devine, J. (1996). Maximum security: The culture of violence in inner-city schools. Chicago: University of Chicago Press. Dodge, K. A., Dishion, T. J., & Lansford, J. E. (2006). Deviant peer influences in programs for youth: Problems and solutions. New York: Guilford. Donovan, J. E., & Jessor, R. (1985). The structure of problem behavior in adolescence and young adulthood. Journal of Consulting and Clinical Psychology, 53, 890 – 904. Donovan, J. E., Jessor, R., & Costa, F. M. (1988). Syndrome of problem behavior in adolescence: A replication. Journal of Consulting and Clinical Psychology, 56(5), 762 – 765. Dowdney, L. (2005). Neither war nor peace: International comparisons of children and youth in organized armed violence. Rio de Janeiro, Brazil: 7 Letras. DuRant, Robert H., et al. 1995. The association between weapon-carrying and the use of violence among adolescents living in or around public housing. Journal of Adolescence 18(5), 579 – 592. Edberg, M. (1992). AIDS risk behavior among runaway youth in the Washington, DC – Baltimore Area. In Report of multi-site runaway risk behavior study. Rockville, MD: National Institute on Drug Abuse. Edberg, M. (1998). Street cuts: Splices from project notebooks and other indelible impressions. Anthropology and Humanism, 23(1), 77 – 82. Edberg, M. (2001). Drug traffickers as social bandits: Culture and drug trafficking in northern Mexico and the border region. Journal of Contemporary Criminology, 17(3), 259 – 277.




Guignon, C. (1993). The Cambridge companion to Heidegger. Cambridge: Cambridge University Press. Hagedorn, J. (2007). Gangs in the global city: Alternatives to traditional criminology. Urbana: University of Illinois Press. Harcourt, Bernard E. 2006. Language of the gun: Youth, crime, and public policy. Chicago: University of Chicago Press. Hawkins, J. D., Catalano, R. F., & Miller, J. Y. (1992). Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: Implications for substance abuse prevention. Psychological Bulletin, 112, 64 – 105. Hawkins, J. D., Herrenkohl, T. I., Farrington, D. P., Brewer, D., Catalano, R. F., Harachi, T. W., & Cothern, L. (2000). Predictors of youth violence. Juvenile Justice Bulletin. Washington, DC: Office of Juvenile Justice and Delinquency Prevention. Hawkins, J. D., & Weis, J. G. (1985). The social development model: An integrated approach to delinquency prevention. Journal of Primary Prevention, 6, 73 – 97. Heidegger, M. (1962). Being and time. New York: Harper & Row. Hirschi, T. (1969). Causes of delinquency. Berkeley: University of California Press. Holland, D., Lachicotte, W., Skinner, D., Cain, C. (1998). Identity and agency in cultural worlds. Cambridge: Harvard University Press. James, W. (1892). Psychology: The briefer course. Notre Dame, IN: Notre Dame University Press. Jessor, R., Donovan, J., & Costa, F. M. (1991). Beyond adolescence: Problem behavior and young adult development. New York: Cambridge University Press. Jessor, R., & Jessor, S. L. (1977). Problem behavior and psychosocial development: A longitudinal study of youth. New York: Academic Press. Jones, Nikki. 2008. Working ‘the code’: On girls, gender, and inner-city violence. Australian & New Zealand Journal of Criminology 41(1), 63 – 83. Karandinos, G. (2010). “You ridin’?” The moral economy of violence in North Philadelphia. Undergraduate thesis, Department of History and Sociology of Science, Program in Health and Societies, University of Pennsylvania. Katz, J. (1988). Seductions of crime: Moral and sensual attractions in doing evil. New York: Basic Books. Kurtines, W. M., Ferrer-Wreder, L., Berman, S. L., Lorente, C. C., Silverman, W. K., & Montgomery, M. J. (2008a). Promoting positive youth development: New directions in developmental theory, methods, and research. Journal of Adolescent Research, 23, 233 – 244. Kurtines, W. M., Montgomery, M. J., Eichas, K., Ritchie, R., Garcia, A., Albrecht, R., Berman, S., Ferrer-Wreder, L., & Lorente, C. C. (2008b). Promoting positive identity development in troubled youth: A developmental intervention science outreach research approach. New directions for identity and intervention [Special issue], Identity: An International Journal of Theory and Research, 8, 125 – 138. Leenhardt, M. (1979). Do Kamo: Person and myth in the Melanesian world. Chicago: University of Chicago Press. Lerner, R. M., Lerner, J. V., Almerigi, J. B., Theokas, C., Phelps, E., Gestsdottir, S., et al.




Padilla, F. M. (1992). The gang as an American enterprise. New Brunswick: Rutgers University Press. Pransky, J. (1991). Prevention: The critical need. Springfield, MO: Burrell Foundation and Paradigm. Sampson, R. J. (2003). The neighborhood context of well-being. Perspectives in Biology and Medicine, 46(3S), S53 – S64. Sampson, R. J., Morenoff, J. D., & Raudenbush, S. (2005). Social anatomy of racial and ethnic disparities in violence. American Journal of Public Health, 95(2), 224 – 232. Sampson R. J., & Wilson, W. J. (1995). Race, crime, and urban inequality. In J. Hagen & R. Peterson (Ed.), Crime and inequality (pp. 37 – 54). Stanford: Stanford University Press. Sandberg, S. (2009a). Gangster, victim or both? The interdiscursive construction of sameness and difference in self-presentations. British Journal of Sociology, 60(3), 523 – 542. Sandberg, S. (2009b). A narrative search for respect. Deviant Behavior, 30(6), 487 – 510. Sandberg, S., & Peterson, W. (2009). Street capital: Black cannabis dealers in a white welfare state. Bristol, UK: Policy. Scales, P. C., Benson, P. L., Leffert, N., and Blyth, D. (2000). Contribution of developmental assets to the prediction of thriving among adolescents. Applied Developmental Science, 4(1), 27 – 46. Scales, P. C., Foster, K. C., Mannes, M., Horst, M. A., Pinto, K. C., and Rutherford, A. (2005). School-business partnerships, developmental assets, and positive outcomes among urban high school students: A mixed-methods study. Urban Education, 40, 144 – 189. Scheper-Hughes, N., & Bourgois, P. (Eds.). (2004). Violence in war and peace: An anthology. Oxford, UK: Blackwell. Schwartz, S. J., Pantin, H., Coatsworth, J. D., & Szapocznik, J. (2007). Addressing the challenges and opportunities for today’s youth: Toward an integrative model and its implications for research and intervention. Journal of Primary Prevention, 28(2), 117 – 144. Search Institute. (1998). Developmental assets: An investment in youth. http://www (accessed December 2012). Singer, M. (2006). A dose of drugs, a touch of violence, a case of AIDS, part II: Further conceptualizing the SAVA syndemic. Free Inquiry for Creative Sociology, 34(1), 39 – 53. Sperber, D. (1996). Explaining culture: A naturalistic approach. Oxford, UK: Blackwell. Sperry, L. L., & Sperry, D. E. (1995). Young children’s presentations of self in conversational narration. In L. Sperry and P. Smiley (Eds.), New directions for child development: Vol. 69. Exploring young children’s concepts of self and other through conversation (pp. 47 – 60). San Francisco: Jossey-Bass. Stewart, Eric A., and Ronald L. Simons. 2006. Structure and culture in African







spatial concentration reduces or contributes to crime (e.g., Clear et al., 2003; Fagan, West, and Holland, 2003; Lynch and Sabol, 2004). These studies examined the reciprocal effects of crime, incarceration, and neighborhood social and economic disadvantage that are bound together in complex neighborhood ecological dynamics. These neighborhood dynamics themselves exert secondary or one-off effects on a range of individual outcomes including crime, employment, school dropout, teenage pregnancy, and drug abuse, often swamping any individual effects (Sampson, Morenoff, and Gannon-Rowley, 2002; Sampson, Morenoff, and Raudenbush, 2005). In some neighborhoods, this racial-spatial concentration may accumulate to produce collective consequences for entire neighborhoods, consequences whose effects are well beyond what we might expect from the aggregation of individual effects of persons within neighborhoods. Several researchers now are examining the effects of this spatial concentration of incarceration, including its effects on social and economic indicia of community life. Recent theoretical and empirical work has focused on the unintended consequences of incarceration not just for individuals or families but for the neighborhoods that experience the highest rates of incarceration (Lynch and Sabol, 2004). Much of this work has focused on the possibility that incarceration may increase neighborhood crime rates (Clear et al., 2003; Fagan, West, and Holland, 2003; Lynch et al., 2001; Lynch and Sabol, 2004). Few (Hagan and Dinovitzer, 1999; Sabol and Lynch, 2003) have looked at the effects of incarceration on the social and economic contexts — human capital, poverty, and family and child well-being — of neighborhoods that are intricately bound up with incarceration and crime (see Crutchfield and Wadsworth, chapter 6 in this volume). Because crime, incarceration, and neighborhood contexts are part of a complex ecological dynamic with reciprocal effects over time, unraveling these influences is a potentially important step in understanding the persistent spatial concentration of incarceration that seems to be orthogonal to local crime rates.4 In this chapter, we take another step in this direction. We analyze data from a panel study of New York City neighborhoods to examine the effects of incarceration on two indicia of the economic well-being of neighborhoods: median household income and human capital. The research setting is New York City in the years from 1985 to 1997, a period




economic behaviors not only of their residents but also of residents of the surrounding areas, through dynamics of diffusion or contagion of neighborhood effects (Reordan et al., 2008; Lee et al., 2008; Reordan and Sullivan, 2004; Grannis, 1998). Neighborhood effects capture the intricate interplay between social structure, social organization, and social control, which combine to influence individual behaviors. Interest in neighborhood effects has produced new research on small-area variations in child development and child maltreatment, domestic violence, teenage sexual behavior and childbearing, school dropout, home ownership, several indicia of health, suicide, social and physical disorder, drug use, and adolescent delinquency (see, for example, Coulton et al., 1995; Miles-Doan, 1998; Crane, 1991; Gould, Wallenstein, and Kleinman, 1990; Gould et al., 1990; Brooks-Gunn et al., 1993; Rowe and Rogers, 1994). Moreover, evidence of the spread of social behaviors from one neighborhood to the next suggests that elements of social contagion may also explain variation in crime rates over time (Fagan and Davies, 2004). Here, we focus not on the neighborhood effects on individuals but instead on the effects of incarceration on the ecology of neighborhoods and their developmental trajectories over time. We assume that neighborhoods (like people) are dynamic entities that change over time and that these transformations are likely to lead to complex outcomes of crime and other indicia of social and economic life. A small number of studies use panel methods to examine these interactions within neighborhoods over time, identifying complex interactions and (nonrecursive) feedback processes between crime and the social dynamics and compositional characteristics of neighborhoods (e.g., Bellair, 2000).5 Some neighborhood-change studies have examined the reciprocal influence of adjacent neighborhoods on crime rates. For example, Taylor and Covington (1988), Morenoff and Sampson (1997), and Heitgerd and Bursik (1987) all identified dynamics in which crime or violence in one area influenced homicide rates in adjacent areas over time. Taylor and Covington examined gentrification as a trigger for crime, while Heitgerd and Bursik used a similar strategy to show that even stable, well-organized communities can have high rates of delinquency when the adjacent neighborhoods experienced rapid racial change. Other studies have identified turning points in neighborhoods




they concluded that incarceration produced more incarceration net of crime and that incarceration was associated with increasing crime. The engine for the growth in incarceration was drug enforcement, which continued to resupply incarceration (Fagan, West, and Holland, 2003, 2005). These dynamics spiraled over time in a reciprocal dynamic that at some tipping point is likely to reach equilibrium. The dynamic becomes self-sustaining and reinforcing and continues even as externalities such as labor market dynamics or population structure undergo significant change, as well as in the face of declining crime rates and receding drug epidemics. In the second study, based on data from a two-wave study of neighborhoods in Tallahassee, Florida, Clear et al. (2003) showed a positive relationship between the rate of releases one year and the community’s crime rates the following year. They showed a dose-response relationship between prison admissions and crime — low rates of prison admissions had nonsignificant effects on local crime rates, moderate prison admission rates produced modest effects on crime, and the neighborhoods with the highest rates had the strongest increase in crime. Provocative as this study may be, it overlooked endogeneity of crime and incarceration that would lead to intercept differences in the neighborhoods at the outset of the panel and weaken the causal claim. That is, higher incarceration rates may simply respond to higher crime rates, or the two may be spuriously related to the factors that produced these intercept differences (i.e., simultaneous equation bias). The Tallahassee study was silent on causal mechanisms, such as incarceration impacts on informal social control or community organization. These mediating mechanisms were an explicit focus in a study by Lynch and Sabol (2004) of crime, incarceration, and social organization in 30 Baltimore communities. Lynch and Sabol examined the effects of neighborhood incarceration rates on community social cohesion and informal social control in the 30 neighborhoods and ultimately the effect of those rates on crime. They tested whether incarceration lessens the capacity of communities to engage in social control, which in turn could increase crime rates. They identified the discretionary component of law enforcement — one of the primary engines of incarceration (see also Fagan, West, and Holland 2003) — through an instrumental variables model to estimate the effects of law enforcement (arrest)




value of sanctions, delegitimizing law and legal actors, further inviting crime, and intensifying the crime-enforcement-incarceration-crime cycle (Fagan and Meares, 2008; Uggen and Manza, 2004). High rates of imprisonment raise questions of the legitimacy of government and potentially undermine incentives to comply with the law (Sherman, 1993; Tyler and Huo, 2002). The racial and neighborhood asymmetry in punishment offers a stark contrast to the claims of legal actors that law is fair and legitimate. If local residents reject the claim that prison sentences are fairly distributed across races and neighborhoods, they may conclude that the policy that produces the unfair distribution is illegitimate (Fagan, 2004).

Incarceration and Neighborhood Economic Well-Being Much of what we know about the adverse effects of incarceration on individuals’ prospects in the legal labor market come from large and small panel studies of former inmates. We were unable to locate studies of the effects of incarceration on the aggregate social or economic wellbeing of neighborhoods as a function of the rates of removal to prison or jail. The panel studies agree that the prospects for stable employment and future earnings of former inmates are dim (Freeman, 1996; Fagan and Freeman, 1999; Western and Pettit, 2000; Western, 2006). As a person’s time spent in prison increases, the subsequent likelihood of disengagement from the legal economy increases (Freeman, 1996; Grogger, 1995; Hagan, 1991). Once out of prison, a criminal record disadvantages lowskill and other workers who are attempting either to enter the labor force or to improve their earnings (Pager, 2003). Western and Beckett’s (1999) study of incarceration and unemployment found that although growing levels of incarceration initially produced lower rates of conventional measures of unemployment, the recycling of these ex-offenders back into the job market with reduced job prospects had the effect of increasing unemployment in the long run. Western (2002) estimates that the earnings loss associated with prison ranges between 10% and 30%, and serving time in prison is also associated with decreased earnings growth. Some studies have looked at the aggregation effects of concentrated




merly performed by the incarcerated family member. In the long run, incarceration will have negative effects on the economic life of the community by reducing the ability of returning inmates to obtain jobs and higher salaries” (2004, 273). This prediction is reinforced when we consider the employer preferences shown by Holzer and colleagues.

Incarceration and Family Integrity One would expect incarceration to be a turning point in the lives of men in several ways that increase their crime risks. Not only are they disadvantaged in the workplace, but their ties to their children and families suffer, eroding an essential form of emotional and social support that has strong effects on criminal activity (Sampson and Laub, 1993; Laub and Sampson, 2003). Recent studies offer evidence that imprisonment damages the ties between incarcerated men and women, their families, and their communities (Hagan and Dinovitzer, 1999, 122; Geller et al., 2009, 2012). These effects further burden the efforts of former inmates to avoid crime once back in their communities but also diminish their capacity to supervise and raise children. Researchers have focused on the fates of families and children, with inferences about communities based on the concentration of incarceration and the aggregation of individual effects. In Random Family, LeBlanc (2003) reports on a social and familial network of Latino families and neighbors. Her ethnography shows how incarceration can weaken families by removing men from existing families, by reducing the supply of marriageable men in the neighborhood, and in turn by attenuating or skewing family formation toward unstable couplings (see also Wilson, 1996). Her work shows the effects of incarceration on the capacities of families as socializing agents for children and on their ability to supervise teenage children. Edin, Nelson, and Paranal (2004) show that incarceration influences the ties between imprisoned men and their children in several ways. In life-history studies with men with low job skills in two cities, they identify a group of men whose ties to their children — ties that were strong prior to incarceration — were disrupted by their imprisonment. Fathers in this group were less able to supervise their children and to maintain




men declines, suppressing the marriage rate. As Wilson explains, “both inner-city black males and females believe that since most marriages will eventually break up and since marriages no longer represent meaningful relationships, it is better to avoid the entanglements of wedlock altogether” (1996, 104). The children of African American incarcerated mothers are far more likely to be placed with another family member or in foster care compared to the children of white incarcerated mothers, even after controlling for differences in social position (Johnson and Waldfogel, 2004, 123). One consequence, then, of higher incarceration rates is strain on the child welfare system. The spatial concentration of incarceration will focus these systemic strains in small social areas with limited foster care resources and supervisory or regulatory capacities.

Incarceration and Local Social Control Recent work with incarcerated males and the “fragile families” they leave behind suggests that incarceration disrupts family ties and social networks, aggravating vulnerabilities to crime through compromises to social control, in turn creating a churning effect on social networks (McLanahan and Sandefur, 1994; McLanahan and Bumpass, 1998; Wildeman, 2010). Social organization and social control are spatially embedded processes that influence neighborhood-level variations in violence (Morenoff, Sampson, and Raudenbush, 2001). Thus, rising and concentrated rates of incarceration not only become a part of the fabric of poor communities, already susceptible to crime, but they compromise the limited forms of social control that poor communities can mount. Informal social control is essential in the regulation of crime (Bursik and Grasmick, 1993). But social control is intricately tied to social structure, supporting citizen activities — social regulation — that can sustain or inhibit crime (Sampson, Raudenbush, and Earls, 1997). When economic conditions are weak, the strains of everyday life can compromise the participation of local residents in social regulation. If these effects extend to neighborhood economic well-being, the strains on residents’ capacity for social control reinforce the crime-incarceration dynamics well observed in other studies.




in addition to prison populations, a dimension of incarceration that has been neglected in much of the research on incarceration. We use the homicide victimization rate as a proxy for the overall crime rate (see Maltz, 1998). We include a series of control variables that capture the dimensions of neighborhood social control and social structure and that themselves are bound up with both incarceration and crime. To estimate temporal effects, we include interactions of time with each of these predictors.

Research Setting and Methods Crime and Incarceration in New York City Trends in crime and incarceration in New York City from 1985 to 1997 provide the backdrop for understanding how incarceration shapes the economic fortunes of neighborhoods. Crime rates rose in New York beginning in 1985, concurrent with the onset of the crack epidemic and the emergence of street drug markets that themselves were flashpoints for violence and other crimes (Fagan, Zimring, and Kim, 1998; Karmen, 2000; Harcourt and Ludwig, 2006). Table 8.1 shows that violent crime rose 29.1% from 1985 to 1990, and the total index crime (i.e., major felonies) rate rose by 18%. Starting in 1991, crime fell sharply, by nearly 50% for index crimes and 46.7% for violent crimes. Incarceration rates rose and fell concurrently with changes in crime rates, though the trajectories were quite different. Prison sentences rose 89% from 1985 to 1990, rising more quickly than the crime rates.9 Prison sentences then declined by 19.2% through 1997, a rate slower than the decline in crime. The steadily increasing rates of prison sentences per reported crime, arrest, and conviction — during periods of both increase and decline in crime — showed the rise in the propensity for incarceration within the criminal justice system in New York City. The effect of these changes in punishment norms was sharp and sustained growth in New York State’s prison population. The state prison population rose from 25,000 in 1985 to 55,000 in 1990 and then to nearly 70,000 in 1997 (Fagan, West, and Holland, 2003). Most — about 70% — of the state’s inmates come from New York City.10 The jail population grew more slowly than did the prison population




increasing proportion of New York State prison admissions: from just 12% of all New York State prison admissions in 1985 to 31% in 1990 to 38% in 1996.13 Because these inmates are likely to serve long sentences under New York’s “predicate felony” laws, drug offenders comprised a growing proportion of the city’s and the state’s incarcerated population (Fagan, West, and Holland, 2003).

Data To estimate the effects of incarceration on neighborhood economic fortunes, we used a longitudinal panel of incarceration, crime, enforcement, and social structure in New York City census tracts for the period from 1985 to 1997 (Fagan, West, and Holland, 2003). We obtained a 25% sample of all individuals sentenced to prison and a 5% sample of all jail sentences for cases with dispositions in New York City for the years 1985, 1987, 1990, 1993, and 1996. This yielded an annual sample of prison sentences of 2,000 to 4,000 individuals and an annual sample of jail sentences of 3,000 to 4,000 individuals. Records of persons admitted to prisons or jails were geocoded by residential address of the incarcerated person. Geocoded cases and crime counts were aggregated to each census tract. Rates of crime and incarceration were then computed for each census tract. We used homicide victimization rates as a proxy of crime generally (Maltz, 1998, 1999) and to account both for base rates of the supply of individuals available for incarceration and the endogeneity of crime with incarceration and neighborhood social organization (Morenoff, Sampson, and Raudenbush, 2001; Fagan and Davies, 2004). Unfortunately, the New York City Police Department does not make available crime data for geographically precise areas such as neighborhoods or census tracts.14 Instead, we used data on homicide victimization from the Office of Vital Statistics of the New York City Department of Health and Mental Hygiene. Deaths are recorded by the Office of the Medical Examiner after classifying injuries as either intentional, accidental, or self-inflicted. Neighborhood rates were estimated by aggregating from individual cases that were geocoded to the census tract using residential address of the victim and by using a population denominator for each year in the time series.15




We used propensity scores of incarceration to identify the “treatment effects” of incarceration on neighborhood economic status. Propensity scores are commonly used to adjust for biases resulting from the nonrandom allocation of subjects to treatment exposures (Rosenbaum and Rubin, 1983; Rosenbaum, 2002). In this case, incarceration is not randomly allocated across the city’s census tracts, and the “dosages” of incarceration similarly reflect nonrandom differences in crime, social structure, and law enforcement (Fagan, West, and Holland, 2003). In this case, propensity scores for both prison and jail are the estimated probability of the allocation of the “treatment” to each neighborhood. Propensity scores thus control for the endogeneity of crime, social structure, and law enforcement, as well as other unobserved confounding variables. We used separate equations to estimate jail and prison propensity scores for each tract in each year of the panel. Following Rubin (1997), we used a set of theoretical predictors to estimate the propensity scores that differed from those used to test the primary research questions. This allows for greater flexibility in model specification than the typical adjustments in regression-based model estimation techniques and more effectively reduces biases resulting from confounding among predictors that is a recurring problem with observational data (Rosenbaum, 2002). Ideally, we would want the functional form of the propensity score analysis to be determined by the data, but in this case, the extreme skew in incarceration rates by tract dictated that we use a log transformation and a linear model. Accordingly, we estimated ordinary least squares regressions for logged jail and prison rates, with predictors including homicide, drug arrests, and a series of social structural factors that are well identified in criminological research on crime and punishment (e.g., Land, McCall, and Cohen, 1990; for a review, see Fagan and Davies, 2004). Following Land et al. (1990), we sorted 18 tract-level variables along seven dimensions — poverty, labor market, segregation, supervision, anonymity, immigration, and housing structure — that characterize the dimensions of concentrated disadvantage articulated in the theoretical and empirical literature linking neighborhood effects with indicia of social adversity and isolation including crime (see, for example, Sampson, Morenoff, and Gannon-Rowley, 2002; Bursik and Grasmick, 1993). For each census year, we used principal components analysis with varimax




the percentage of female-headed households with children under 15, and the ratio of youths to adults. The general model is _ _ _ (Yit–θYi ) = (1–θ)a + (Xit–θXi )b + [(1–θ)α1 + (η–θηi )], where Yit is economic measure of each census tract i for each time period t, Yi is the mean of Y over time for each tract, and X is a vector including the incarceration propensity scores and other predictors. We include time as both a fixed and random effect: time is included as a random effect to account for the panel structure of the data and as a fixed effect to account for the specific year within the panel. We include an interaction term of time by each predictor to estimate their specific longitudinal effects. In this form, the main effect represents the average effect of the predictor across the time series, with the interaction with time as the longitudinal effect. We focus on the latter to identify the cumulative longitudinal effects of incarceration.

Results Patterns and Trends We begin by showing the concentration of prison admissions and their relationship to income and human capital in New York City neighborhoods. Figures 8.1A and 8.1B compare prison admissions by median household income for two periods: 1985 – 1990 and 1993 – 1996, periods of increasing and then declining crime in New York City; figures 8.2A and 8.2B similarly compare incarceration with human capital. The patterns show the strong inverse correlation for each economic indicator with the rate of prison admissions: prison admissions are concentrated in neighborhoods with the lowest incomes and the lowest human capital. The figures also show the stability of incarceration by neighborhood during two distinctly different crime eras. Despite strong crime declines in New York City, prison admissions were concentrated in the same neighborhoods. Tables 8.2 and 8.3 show change over time in household income and human capital using quartiles to group neighborhoods according to their concentration of prison incarceration.




Table 8.2. Median Household Income by Incarceration Quintiles, 1986–1997, New York City Census Tracts (Means, Standard Deviations) Year Quintile






No Events

28,523 (10535) 24,022 (8195) 21,706 (8961) 17,709 (8095) 14,832 (7373) 24,675 (10815)

32,740 (12241) 28,917 (10597) 25,186 (10888) 21,058 (9550) 17,621 (8977) 27,972 (12527)

38,354 (14320) 34,389 (12733) 31,171 (10946) 24,704 (10776) 18,790 (10715) 32,218 (14584)

41,890 (16197) 36,720 (12280) 33,250 (11361) 26,162 (10988) 20,066 (9816) 35,012 (15854)

44,039 (16568) 40,047 (17943) 34,654 (12197) 28,715 (13908) 22,198 (14212) 37,821 (17526)

1 2 3 4 Total

Source: Bureau of the Census, STF 3A, Interpolated for Reconciled Census Tracts 1980–2000

Table 8.3. Human Capital (Factor Score) by Incarceration Quintiles, 1986–1997, New York City Census Tracts (Means, Standard Deviations) Quintile



Year 1991



No Events

0.261 (0.854) 0.288 (0.931) –0.143 (0.896) –0.583 (0.875) –0.940 (0.840) 0.001 (0.966)

0.283 (0.870) 0.317 (0.868) –0.096 (0.921) –0.485 (0.869) –0.870 (0.929) 0.000 (0.978)

0.316 (0.841) 0.409 (0.898) 0.051 (0.786) –0.432 (0.918) –1.049 (0.844) 0.000 (0.987)

0.335 (0.800) 0.334 (0.893) 0.037 (0.845) –0.483 (0.859) –1.053 (0.796) 0.000 (0.964)

0.302 (0.826) 0.253 (0.933) –0.063 (0.861) –0.477 (0.918) –0.958 (0.881) 0.000 (0.968)

1 2 3 4 Total

Source: Bureau of the Census, STF 3A, Interpolated for Reconciled Census Tracts 1980–2000

Facing page: Top, Fig. 8.1, A, 1985 – 1990 median household income; B, 1993 – 1996 median household income Bottom, Fig. 8.2, A, 1985 – 1990 human capital; B, 1993 – 1996 human capital




neighborhoods with no incarceration and then a slight decline. For the first quartile, those neighborhoods with the lowest incarceration rates, human capital remained stable over time. In the second and third quartiles, human capital increased through 1991 before declining slightly through 1997. The trend in the fourth quartile differed: human capital rose slightly before declining sharply and then rising slightly again in the last period in the study interval. Incarceration Effects: Model Estimation Models were estimated in four ways to identify more specifically the effects of both jail and prison on neighborhood economic status. Models for jail and prison were estimated separately. Models with both jail and prison were then estimated to examine their additive effects on economic measures. The fourth model examined their conditional effects by including an interaction term that combined jail and prison admissions. We included race-specific measures of neighborhood demography. Each set of models included first-order interactions of each predictor with time to examine the effects over time of incarceration and the other predictors.18 Descriptive statistics for the variables used in the models are shown in table 8.4 and also in the appendix. Table 8.4. Descriptive Statistics Median household income Population > 15 % African American % nonwhite Hispanic Prison rate* Jail rate* Drug arrest rate* Homicide victimization rate† Poverty/inequality (factor) Human capital (factor) Segregation (factor) Social control (factor) Anonymity (factor) Immigration (factor) Housing structure (factor) * Rate per 10,000 persons † Rate per 1,000 persons


Std. Dev.



31,541 3018 26.2 22.6 2.4 7.9 9.6 0.24 0.0 0.0 0.0 0.0 0.0 0.0 0.0

15,208 2361 32.1 21.9 4.3 18.2 19.9 0.36 1.0 1.0 1.0 1.0 1.0 1.0 1.0

4,757 16 0.0 0.0 0.0 0.0 0.0 0.0 –3.1 –5.1 –2.8 –5.2 –3.4 –1.8 –3.4

177,088 34079 100.0 100.0 114.1 421.1 619.8 6.5 3.1 3.0 1.7 4.9 8.9 4.5 13.7




through the propensity score estimations for jail and prison, perhaps owing to the scale of race effects on neighborhood economic status relative to highly variable incarceration rates by census tract. In the conditional model, the coefficients for jail and prison are still negative and significant, while the interaction term is positive and significant. However, there is little change in the explanatory power of the additive and conditional models. There is no change in the R2 between the jail-only model and the additive model and only a very modest gain in explained variance in the conditional model. In these data, then, there is little evidence of interactions between jail and prison: the effects appear to be neither conditional nor interdependent, nor are they additive in their effects on neighborhood incomes. The opposing effects of jail and prison are not surprising as a matter of crime control and neighborhood ecology. Admissions to jail and prison are processes that occur concurrently, though they are perhaps sustained by distinct patterns of policing and enforcement. In New York, jail admissions result from enforcement of quality-of-life crimes and low-level misdemeanors, consistent with order-maintenance policing strategies (Harcourt, 2001; Greene, 1999), whereas drug enforcement and other anticrime activities are more likely to produce prison-generating felony arrests (Fagan, West, and Holland, 2003). Nevertheless, the models estimating the propensity scores suggest that the two incarceration processes are concentrated in similar, if not identical, neighborhoods. In fact, their bivariate correlation averaged over the study interval is .914. Including both prison and jail in the same model (table 8.5, columns 3 and 4) may conceptually more accurately capture the dynamics of incarceration within neighborhoods than is portrayed by separate models, but the coefficients are almost indistinguishable from other specifications. The additive and conditional models produce similar results, and in both estimations, the effects for jail and prison each decrease a community’s economic outlook. Incarceration Effects on Human Capital The effects of prison and jail admissions on human capital follow a similar pattern. The main effects of prison and jail models (table 8.6, columns 1 and 2) are significant and negative. However, neither jail nor




and psychological well-being of those left behind. Only a few studies have attended to the collective fates of neighborhoods with elevated incarceration rates, and most of these projects have examined how incarceration patterns contribute to the endogenous and spiraling relationship between crime and incarceration. Given prison’s unique and heavy individual costs, the emphasis in incarceration research has been primarily on prisons, with little attention to frequent though shorterterm stays in local jails. Here, we address each of these dimensions of research on incarceration. Loosely Coupled Enforcement We find distinct, yet consistent, effects for prisons and jails, suggesting that these separate processes based on loosely coupled law enforcement priorities and penal strategies have a distinct impact on communities in New York City. Heterogeneous policing regimes in New York City have created separate streams of prisoners eligible for prison and jail. During much of the study period, prison populations were driven by street drug enforcement aimed at low-level dealers (Fagan, West, and Holland, 2003; Smith et al., 1992; Jacobson, 2005). This enforcement strategy was carried out by elite police units such as the Tactical Narcotics Teams, known locally as TNT (Smith et al., 1992; Fagan, 1994; Kleiman, 1992). TNT units were assigned to broad areas of the city rather than to specific precincts. They made tens of thousands of felony drug arrests each year beginning in 1988 and continuing through 1992, heavily populating prisons with felony drug offenders, changing the offense composition of prisons (Letwin, 1990; Herman, 1999; Fagan, West, and Holland, 2003). But jail populations were driven by enforcement of misdemeanor laws, including lesser drug crimes and local enforcement of incivilities and minor misdemeanors. Uniformed patrol officers assigned to precincts, without the organizational attention or status that was conferred on the specialized street drug details, were the front lines in this less visible but persistent enforcement strategy. However, it would seem that the jail removals have had the more profound and lasting impact on communities. Only for the last of the waves in this panel — 1996 — had New York’s Order Maintenance Policing (OMP) been implemented, which switched




and prison have lasting effects on incomes, effects that persist over time. Incarceration in prison seems to give rise not only to more imprisonment (Fagan, West and Holland, 2003) but also to lower incomes. Jail serves not only as a pathway to prison but also as a profound destabilizing influence on communities. The reinforcing spiral of poverty, crime, and incarceration describes what Sampson and Morenoff (2006) characterize as a poverty trap (see also Fagan, 2008). High rates of neighborhood imprisonment can have multiple effects that sustain the downward pressure on local incomes: tainting nonoffender residents who seek jobs as crime risks, burdening returning inmates with reduced prospects in the workplace, and reducing the attractiveness of such neighborhoods for economic or housing investments. The effect of jail on incomes is suggestive of the increasing reliance on the use of jail as an enforcement strategy across New York City’s residential neighborhoods. Places with high jail rates are characterized if not by high rates of low-level crime, then by at least low-level social disorder of the kind that attracts law enforcement attention and a kind of enforcement that removes people for short but disruptive periods. The repetitive high rates of removal to jail of low-level offenders from poor communities, followed by their quick release, create a churning effect on neighborhoods that destabilizes social control and instills a sense of chaos and disorder more typical of a version of broken windows theory (Wilson and Kelling, 1982; Kelling and Cole, 1996; Livingston, 1997; Maple and Mitchell, 2000; Waldeck, 2000; Harcourt, 2001). Also, jail populations often are socially and psychologically troubled, and their interrupted presence in struggling communities is unlikely to aid the efforts of residents in those areas to develop economically or to become socially cohesive (Jacobson, 2005). Ex-inmates also have dim prospects in the workplace, but their diffusion across neighborhoods dilutes their concentrated effects on local incomes. Race and ethnicity effects on income work in unanticipated ways and require further analyses. While prison and jail inmates come from predominantly poor and segregated communities, once incarceration is factored out of the equation, the difference in the proportions of African Americans was not related to incomes, but there was a lasting negative influence on neighborhood fortunes. We observed greater effects for neighborhoods with higher proportions of Hispanics, notwithstanding




private investment capital (Sen, 2011). Microinvestment banks, as economic innovators, fill this gap and provide startup opportunities where other support is harder to access. Some governments also have experimented with this model to spur economic development and job creation in economically isolated areas (Deutsch, 2006). Locating these businesses in the neighborhoods where incarceration rates are higher could provide access to jobs for former inmates and remove some of the barriers to employment that seem to sustain these poverty traps (see, generally, Lyons and Pettit, 2011). Housing issues also intersect with economic development in areas characterized by high rates of crime and incarceration (Schwartz, Susin, and Voicu, 2003). New York experienced a housing and real estate boom starting in the mid-1990s that was concurrent both temporally and spatially with crime declines that persisted through much of the following decade (Fagan and Davies, 2007; Fagan, 2008). At the same time that residential real estate rose in value, so too did the value of commercial real estate, suggesting potential demand for sustainable small businesses in the neighborhoods with high crime and incarceration rates. Abandoned or dilapidated housing was condemned and then transformed into affordable low-income units and made available to local families who qualified for housing assistance (van Ryzin and Genn, 1999). Beginning in 1990, at the outset of New York City’s crime decline, the city created nearly 250,000 in rem housing units from abandoned properties in the city’s poorest and highest crime and incarceration neighborhoods. These units were made available to families qualifying under a variety of low-income and affordable housing programs (van Ryzin and Genn, 1999; Fagan, 2008). The increase in affordable housing in high-crime areas such as the South Bronx, which transformed during that time into a stable, low-crime, and economically diverse neighborhood (Fagan, 2008), suggests the potential for housing to intervene in neighborhoods to disrupt the poverty traps that link crime, incarceration, and economic disadvantage (Sampson and Morenoff, 2006). Human capital may also be sensitive to these types of economicdevelopment policies, as well as to the linkage between education and crime-control policies. One connection may be in the use of Order Maintenance policing strategies in schools, which results in high rates of suspension and expulsion for both violations of school rules and



Notes 1. This research was supported by Grant 85-00-11 from the Russell Sage Foundation, Future of Work Program. All opinions are solely those of the authors, as are any errors. Additional support was provided by the Robert Wood Johnson Foundation, Substance Abuse Policy Research Program. We are grateful to the New York State Division of Criminal Justice Services, the New York State Department of Labor, the New York City Police Department, and the New York City Department of Health for generously supplying data for this research. Tamara Dumanovsky helped conceptualize and launch the project and supervised the assembly of the data sets. Jan Holland prepared the maps and supervised the geocoding of incarceration and arrest records. Steven Glickman, Nicole Mutter, and Carolyn Pinedo provided excellent research assistance. 2. Disenfranchisement disproportionately and severely affects African American males, consistent with their disproportionate presence in the incarcerated population: of the 3.9 million American felons who were disenfranchised in 1999, nearly 1.4 million were African American males, representing 13% of all black males (Maurer, 2006). 3. For example, neighborhood disadvantage may invite closer surveillance by law enforcement, well in excess of levels of surveillance and enforcement that would be predicted by crime rates alone (Fagan and Davies, 2000, 2002), increasing incarceration risks relative to crime rates. These reciprocal patterns of crime, enforcement, and social risk sustain the elevated rates of incarceration and appear to do so even when crime rates decline (Fagan, West, and Holland, 2003). 4. See, for example, Fagan, West, and Holland, 2005, showing the persistence over time of incarceration rates that are independent of local crime rates, even after accounting for the endogeneity of crime, incarceration, and social structural disadvantage in New York City neighborhoods. 5. Physical and social deterioration is a persistent theme of neighborhood change in several studies (Taub, Taylor, and Dunham, 1984; Schuerman and Kobrin, 1986; Harrell and Gouvis, 1994). Deterioration often cued citizens to leave previously stable areas on the basis of changes in their subjective evaluation of the likelihood of crime affecting them personally. 6. That is, they computed the portion of the rate of drug arrests in each neighborhood that was not explained by the index crime rate. 7. The positive effect of incarceration on informal social control may, at first glance, be unexpected. Lynch and Sabol suggest that changes (increases) in incarceration rates encourage informal social control through mechanisms such as fear reduction. Because they failed to find that incarceration promotes prosocial interactions among residents, they suggest that the incarceration – informal social control linkage operates through individuals: “Residents may see or know of persons being incarcerated for crime, and this may increase their confidence in engaging in informal social control. They may feel that the ‘bad guys’ are gone


8. 9.


11. 12.

13. 14.



backfiring. Or a burglary could simply be a missing personal item that was later recovered. Once verified, complaints were entered into the city’s crime counts, but for unstated reasons, the geographical coordinates of the crime location were not carried forward or aggregated. Although using residential address in lieu of event location may distort the spatial estimates for violent events, we based this decision on prior work showing the close proximity of homicide events to the residences of victims. See, for example, Fagan and Wilkinson, 1998. We estimated models with dummies for the two highest quartiles as “high incarceration” tracts and with the top quartile as “very high incarceration.” Models were estimated substituting these indicia of prison for the actual prison rates. The results were robust to these specifications. Separate models were estimated with either fixed and random effects, and a Hausman test was employed to test for the choice between the two kinds of models. Models with fixed versus random effects produced similar results, though coefficients varied in magnitude but not significance. Alternate specifications included race-specific incarceration measures and models that included a dummy for whether the neighborhood was in the highest quartile of incarceration rates to identify whether effects were concentrated at the extremes of the distribution of jail or prison rates. The results were unchanged.

References Baumer, Eric, Janet L. Lauritsen, Richard Rosenfeld, and Richard Wright. 1998. “The Influence of Crack Cocaine on Robbery, Burglary, and Homicide Rates: A CrossCity, Longitudinal Analysis.” Journal of Research in Crime and Delinquency 35 (3): 316 – 340. Becker, Gary S. 1994. Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education. 3rd ed. Chicago: University of Chicago Press. Belenko, Steven, Jeffrey Fagan, and Tamara Dumarovsky. 1994. “The Impact of Special Drug Courts on Recidivism of Felony Drug Offenders.” Justice System Journal 17 (1): 53 – 82. Bellair, Paul E. 2000. “Informal Surveillance and Street Crime: A Complex Relationship.” Criminology 38 (1): 137 – 169. Beveridge, Andrew A. 2008. “A Century of Harlem in New York City: Some Notes on Migration, Consolidation, Segregation, and Recent Developments.” City and Community 7:358 – 365. Bowling, Benjamin. 1999. “The Rise and Fall of New York Murder: Zero Tolerance or Crack’s Decline?” British Journal of Criminology 39 (4): 531 – 554. Bratton, William, and Peter Knobler. 1998. The Turnaround: How America’s Top Cop Reversed the Crime Epidemic. New York: Random House.




Incarceration as Potential Turning Points in the Criminal Careers of Unskilled Men.” In Mary Pattillo, David Weiman, and Bruce Western (eds.), Imprisoning America: The Social Effects of Mass Incarceration, 46 – 75. New York: Russell Sage Foundation Press. Fagan, Jeffrey. 1994. “Do Criminal Sanctions Deter Drug Offenders?” In Doris MacKenzie and Craig Uchida (eds.), Drugs and Criminal Justice: Evaluating Public Policy Initiatives, 188 – 214. Newbury Park, CA: Sage. Fagan, Jeffrey. 2004. “Crime, Law and Community: Dynamics of Incarceration in New York City.” In Michael Tonry (ed.), The Future of Imprisonment, 27 – 60. New York: Oxford University Press. Fagan, Jeffrey. 2008. “Crime and Neighborhood Change.” In Arthur S. Goldberger and Richard Rosenfeld (eds.), Understanding Crime Trends, 81 – 126. Washington, DC: National Academy Press. Fagan, Jeffrey, and Garth Davies. 2000. “Street Stops and Broken Windows: Terry, Race, and Disorder in New York City.” Fordham Urban Law Journal 28:457 – 504. Fagan, Jeffrey, and Garth Davies. 2002. The Effects of Drug Enforcement on the Rise and Fall of Homicides in New York City, 1985 – 95. Final Report, Grant 031675, Substance Abuse Policy Research Program, Robert Wood Johnson Foundation. Fagan, Jeffrey, and Garth Davies. 2004. “The Natural History of Neighborhood Violence.” Journal of Contemporary Criminal Justice 20 (2): 127 – 147. Fagan, Jeffrey, and Garth Davies. 2007. “The Political Economy of the Crime Decline in New York City.” Paper presented at the annual meeting of the American Association for the Advancement of Science, San Francisco, February. Fagan, Jeffrey, and Richard B. Freeman. 1999. “Crime and Work.” In Michael Tonry (ed.), Crime and Justice, vol. 25, A Review of Research, 113 – 178. Chicago: University of Chicago Press. Fagan, Jeffrey, Amanda Geller, Garth Davies, and Valerie West. 2010. “Street Stops and Broken Windows Revisited: Race and Order Maintenance Policing in a Safe and Changing City.” In Stephen K. Rice and Michael D. White (eds.), Exploring Race, Ethnicity and Policing: Essential Readings, 309 – 348. New York: NYU Press. Fagan, Jeffrey, and Tracey L. Meares. 2008. “Punishment, Deterrence and Social Control: The Paradox of Punishment in Minority Communities.” Ohio State Journal of Criminal Law 6:173 – 229. Fagan, Jeffrey, Valerie West, and Jan Holland. 2003. “Reciprocal Effects of Crime and Incarceration in New York City Neighborhoods.” Fordham Urban Law Journal 30:1551 – 1602. Fagan, Jeffrey, Valerie West, and Jan Holland. 2005. “Neighborhood, Crime, and Incarceration in New York City.” Columbia Human Rights Law Review 36:71 – 108. Fagan, Jeffrey, and Deanna L. Wilkinson. 1998. “Guns, Youth Violence and Social Identity in Inner Cities.” In Michael Tonry and Mark H. Moore (eds.), Crime and Justice, vol. 24, Youth Violence, 373 – 456. Chicago: University of Chicago Press. Fagan, Jeffrey, Deanna L. Wllkinson, and Garth Davies. 2007. “Social Contagion of Violence.” In Daniel J. Flannery, Alexander T. Vazsonyi, and Irwin D. Waldman




Harcourt, Bernard E. 2001. Illusion of Order: The False Promise of Broken Windows Policing. Cambridge: Harvard University Press. Harcourt, Bernard E., and Jens Ludwig. 2006. “Broken Windows: New Evidence from New York City and a Five-City Social Experiment.” University of Chicago Law Review 73:271 – 320. Harrell, Adele, and Caterina Gouvis. 1994. Predicting Neighborhood Risk of Crime. Washington, DC: Urban Institute. Heitgerd, Janet L., and Robert J. Bursik. 1987. “Extracommunity Dynamics and the Ecology of Delinquency.” American Journal of Sociology 92 (4): 775 – 787. Herman, Susan. 1999. “Measuring Culpability by Measuring Drugs? Three Reasons to Evaluate the Rockefeller Drug Laws.” Albany Law Review 63:777 – 798. Holzer, Harry J., Steven Raphael, and Michael Stoll. 2004. “Will Employers Hire ExOffenders? Employer Preferences, Background Checks, and Their Determinants.” In Mary Patillo, David Weiman, and Bruce Western (eds.), Imprisoning America: The Social Effects of Mass Incarceration, 205 – 243. New York: Russell Sage Foundation Press. Jacobson, Michael. 2005. Downsizing Prisons: How to Reduce Crime and End Mass Incarceration. New York: NYU Press. Johnson, Elizabeth I., and Jane Waldfogel. 2002. “Parental Incarceration: Recent Trends and Implications for Child Welfare.” Social Service Review 76 (3): 460 – 479. Johnson, Elizabeth I., and Jane Waldfogel. 2004. “Children of Incarcerated Parents: Multiple Risks and Children’s Living Arrangements.” In Mary Pattillo, David Weiman, and Bruce Western (eds.), Imprisoning America: The Social Effects of Mass Incarceration, 97 – 131. New York: Russell Sage Foundation Press. Karmen, Andrew. 2000. New York Murder Mystery: The True Story behind the Crime Crash of the 1990s. New York: NYU Press. Kasinitz, Philip, John H. Mollenkopf, Mary C. Waters, and Jennifer Holdaway. 2008. Inheriting the City: Children of Immigrants Come of Age. New York: Russell Sage Foundation Press. Katz, Lawrence, Steven D. Levitt, and Ellen Shustorovich. 2003. “Prison Conditions, Capital Punishment, and Deterrence.” American Law and Economics Review 5:318 – 343. Kelling, George, and Catherine Cole. 1996. Fixing Broken Windows. New York: Free Press. Ketcham, Christopher. 2002. “Roach Motel.” Salon, October 17. com/mwt/feature/2002/10/17/jail_time/index_np.html. Kleiman, Mark. 1992. Against Excess: Drug Policy for Results. New York: Basic Books. Land, Kenneth, Patricia McCall, and Lawrence Cohen. 1990. “Structural Covariates of Homicide Rates: Are There Any Invariances across Time and Space?” American Journal of Sociology 95:922 – 963. Laub, John H., and Robert J. Sampson. 2003. Shared Beginnings, Divergent Lives: Delinquent Boys to Age 70. Cambridge: Harvard University Press.




Dynamics of Neighborhood Transition: Chicago, 1970 – 1990.” Social Forces 76 (1): 31 – 64. Morenoff, Jeffrey D., Robert J. Sampson, and Stephen Raudenbush. 2001. “Neighborhood Inequality, Collective Efficacy, and the Spatial Dynamics of Urban Violence.” Criminology 39:517 – 553. Murray, Joseph, David P. Farrington, Ivana Sekol, and Rikke F. Olsen. 2009. “Effects of Parental Imprisonment on Child Antisocial Behaviour and Mental Health: A Systematic Review.” Campbell Systematic Reviews 4. Myers, Samuel L., Jr. 2000. “The Unintended Impacts of Sentencing Guidelines on Family Structure.” Paper presented at the annual meeting of the American Sociological Association, Washington, DC, August. Nagin, Daniel S., Francis T. Cullen, and Cheryl Lero Jonson. 2009. “Imprisonment and Reoffending.” In Michael Tonry (ed.), Crime and Justice, vol. 38, A Review of Research, 115 – 200. Chicago: University of Chicago Press. Ousey, Graham C., and Matthew R. Lee. 2002. “Examining the Conditional Nature of the Illicit Drug Market – Homicide Relationship: A Partial Test of the Theory of Contingent Causation.” Criminology 40:73 – 102. Pager, Devah. 2003. “The Mark of a Criminal Record.” American Journal of Sociology 108 (5): 937 – 975. Pager, Devah. 2007. Marked: Race, Crime, and Finding Work in an Era of Mass Incarceration. Chicago: University of Chicago Press. Petersilia, Joan. 2003. When Prisoners Come Home: Parole and Prisoner Reentry. New York: Oxford University Press. Pettit, Becky, and Christopher J. Lyons. 2007. “Status and the Stigma of Incarceration: The Labor Market Effects of Incarceration by Race, Class, and Criminal Involvement.” In Shawn Bushway, Michael Stoll, and David Weiman (eds.), Barriers to Reentry? The Labor Market for Released Prisoners in Post-industrial America, 203 – 226. New York: Russell Sage Foundation Press. Pettit, Becky, and Bruce D. Western. 2004. “Mass Imprisonment and the Life Course: Race and Class Inequality in U.S. Incarceration.” American Sociological Review 69 (2): 151 – 169. Rashbaum, William. 2000. “Police Suspend Extra Patrols for 10 Days.” New York Times, Oct. 12, B1. Raudenbush, Stephen W., Anthony Bryk, and Richard Congdon. 2002. Hierarchical Linear Modeling. Thousand Oaks, CA: Sage. Reordan, Sean F., Chadd R. Farrell, Stephen A. Matthews, Daniel O’Sullivan, and Kenneth Bischoff. 2008. “The Geographical Scale of Metropolitan Racial Segregation.” Demography 45:489 – 514. Reordan, Sean F., and Daniel O’Sullivan. 2004. “Measures of Spatial Segregation.” Sociological Methodology 34:121 – 162. Rose, Dina R., and Todd R. Clear. 1998. “Incarceration, Social Capital, and Crime: Implications for Social Disorganization Theory.” Criminology 36:441 – 480. Rosenbaum, Paul R. 2002. Observational Studies. 2nd ed. New York: Springer-Verlag.




Hierarchical Models, and Individual Growth Models.” Journal of Educational and Behavioral Statistics 24 (4): 323 – 355. Singer, Judith D., and John B. Willett. 2003. Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York: Oxford University Press. Smith, Michael E., Michele Sviridoff, Susan Sadd, Ric Curtis, and Randy Grinc. 1992. The Neighborhood Effects of Street-Level Drug Enforcement – Tactical Narcotic Teams in New York: An Evaluation of TNT. New York: Vera Institute. Spelman, William. 2000. “What Recent Studies Do (and Don’t) Tell Us about Imprisonment and Crime.” Crime and Justice 27:419 – 494. Spitzer, Elliott. 1999. “The New York City Police Department’s ‘Stop and Frisk’ Practices.” Office of the New York State Attorney General. Available at http://www.oag. Taub, Richard P., Garth D. Taylor, and Jan D. Dunham. 1984. Paths of Neighborhood Change: Race and Crime in Urban America. Chicago: University of Chicago Press. Taylor, Ralph B., and Jeanette Covington. 1988. “Neighborhood Changes in Ecology and Violence.” Criminology 26:553 – 589. Travis, Jeremy. 2005. But They All Come Back: Facing the Challenges of Prisoner Reentry. Washington, DC: Urban Institute Press. Tyler, Tom R., and Yuen J. Huo. 2002. Trust in the Law: Encouraging Public Cooperation with the Police and Courts. Vol. 5. New York: Russell Sage Foundation Press. Uggen, Christopher, Angela Behrens, and Jeff Manza. 2005. “Criminal Disenfranchisement.” Annual Review of Law and Social Science 1:307 – 322. Uggen, Christopher, and Jeff Manza. 2002. “Democratic Contraction? Political Consequences of Felon Disenfranchisement in the United States.” American Sociological Review 67:777 – 803. Uggen, Christopher, and Jeff Manza. 2004. “Lost Voices: The Civic and Political Views of Disenfranchised Felons.” In Mary Patillo, David Weiman, and Bruce Western (eds.), Imprisoning America: The Social Effects of Mass Incarceration, 165 – 204. New York: Russell Sage Foundation Press. van Ryzin, Gregg G., and Andrew Genn. 1999. “Neighborhood Change and the City of New York’s Ten-Year Housing Plan.” Housing Policy Debate 10 (4): 799 – 838. Visher, Christy A., and Jeremy Travis. 2003. “Transitions from Prison to Community: Understanding Individual Pathways.” Annual Review of Sociology 29:89 – 113. Waldeck, Sarah. 2000. “Cops, Community Policing, and the Social Norms Approach to Crime Control: Should One Make Us More Comfortable with the Others?” Georgia Law Review 34:1253 – 1297. Western, Bruce D. 2002. “The Impact of Incarceration on Wage Mobility and Inequality.” American Sociological Review 64:526 – 546. Western, Bruce D. 2006. Punishment and Inequality in America. New York: Russell Sage Foundation Press Western, Bruce D., and Katherine Beckett. 1999. “How Unregulated Is the U.S. Labor Market? The Penal System as a Labor Market Institution.” American Journal of Sociology 104:1030 – 1060.







how macroeconomic factors influence the developing child across multiple contexts that are differentially salient at different developmental stages and can increase or decrease risk for violence. For young children, the most relevant developmental contexts are families and communities, with families clearly embedded in communities. As children enter school, the school and peer context become important as well. During adolescence, the influence of peers in shaping deviant behavior increases, and more direct macroeconomic impacts linked to opportunities for employment and productive engagement may come into play. The purpose of this chapter is to examine potential linkages between macroeconomic conditions, developmental processes, and youth violence from the prenatal period through adolescence. Much of the developmental literature has examined biological, psychological, and social processes related to aggression and violence, with some studies considering how these processes are impacted by macroeconomic conditions. Most typically the focus has been on how chronic and extreme poverty impact development, although some studies have examined other factors such as economic downturns (Elder, 1999) and unemployment or underemployment (Yoshikawa, Weisner, & Lowe, 2006). In this chapter, the focus is on macroeconomic conditions most likely to disrupt relevant contextual processes at different developmental stages and, consequently, to increase risk for aggression and violence. This includes a discussion of how biological development can be compromised as well as how these processes impact psychological development and the learning of aggression and violence over time. Emphasis is placed on early neurological development and the child’s emerging cognitive system, including social-information-processing skills, attitudes, and beliefs that have been associated with higher levels of childhood aggression and youth violence (Crick & Dodge, 1994). Specification of these risk factors associated with violence and linked to macroeconomic conditions also provides directions for policy and interventions to prevent youth violence under adverse economic circumstances. A developmental perspective also provides a unique lens to examine inconclusive or inconsistent evidence regarding the effects of economic downturns on youth violence. Although there is general support for a relation between economic downturns and increases in some indices of youth violence, even the chapters in this book reveal different




whose troubled behavior is evident early in development and persists over time (Moffitt, 2003). Still, within each developmental stage, macroeconomic factors are most likely to compromise children’s development and possibly increase risk for youth violence via their effects on key developmental contexts and processes. Let us now turn to a review of risk factors likely to be impacted by macroeconomic conditions across specific developmental stages: the prenatal period through age 5; the early school years (ages 6 – 11); adolescence (ages 12 – 18). The Prenatal Period through Age 5 During this developmental stage, communities and families are the most relevant context for child development. A number of scholars have proposed that community conditions related to low socioeconomic status and/or poverty primarily influence child development via their influence on parental behavior (e.g., Linver, Brooks-Gunn, & Kohen, 2002; Mistry, Vandewater, Huston, & McLoyd, 2002). Limited resources and services in a community can translate into blunted opportunities for adequate prenatal and postnatal care and compromised parenting due to high levels of family stress. Direct effects of community poverty that can operate independent of their influence on families also are evident, including increased exposure to environmental toxins, lack of opportunities for early enrichment, and psychological trauma linked to witnessing community violence. Children growing up in poor families living in economically disadvantaged communities thus experience a double whammy of community-level stressors that affect both children and families and that parents are unlikely to be able to buffer adequately. Each of these factors can increase risk for childhood aggression and youth violence. Maternal behavior has direct bearing on children’s development during the prenatal period, and a number of maternal characteristics and behaviors can lead to higher levels of biological and neuropsychological vulnerability associated with later aggression. Chronic poverty has long been associated with inadequate prenatal care and a higher likelihood of fetal exposure to toxic agents, possibly leading to impaired brain development, lower birth weight, nutritional deficiencies, and greater risk for assorted neurological problems. Even birth complications associated




children who are not engaged in securely attached relationships with parents or caregivers may escalate their angry and aggressive behavior in order to elicit a response from them. Parents who are ill equipped to manage this behavior, whether due to a history of disadvantage or immediate stressors, may respond with increased punishment and retaliation. This pattern of insecure attachment can lead to problematic internal working models reflecting cognitive representations of relationships that may compromise the developing child’s ability to develop trusting and stable friendships and long-term intimate relationships later in life (Bowlby, 1973). In addition to problems associated with weak and insecure attachment bonds, harsh parenting and corporal punishment can increase the likelihood of child aggression and youth violence. The use of corporal punishment is greatest toward young children age 5 and under and among low-socioeconomic-status parents (Straus & Stewart, 1999). There are several reasons for the relation between early corporal punishment and subsequent aggression and violence. First, corporal punishment communicates to the child that aggression and acting-out behaviors are normative, acceptable, and effective ways to gain compliance (Bandura, 1986). Second, corporal punishment often leads children to avoid the disciplinary figure, reducing available opportunities for parents to direct and influence their child (which already may be limited under adverse economic conditions). Third, corporal punishment leads children to be hypervigilant to hostile cues, to attribute hostile intent to others, to access more aggressive potential responses, and to view aggression as a way to attain social benefits (Dodge, Pettit, McClaskey, & Brown, 1986). Fourth, corporal punishment can motivate children to avoid future punishment but does not teach children the responsibility to behave independently in morally and socially acceptable ways (Gershoff, 2002). Thus, not only are children developing characteristic patterns of behavior that include aggression (and facilitate later violence), but they are also forming early patterns of social cognition — and what they are learning is that aggression and violence are integral parts of social relationships. The evidence linking physical abuse, aggression, and later violence also is compelling (Luntz & Widom, 1994). In many cases, this abuse begins as corporal punishment and is then taken to the extreme. In




underdevelopment of the cortex. Because the cortex plays a major role in regulating the lower parts of the central nervous system, increased cortical capacity should be expected to decrease propensity for violence over the course of development (Perry, 1997). The importance of early experience for the child’s neurological and psychological development cannot be understated. Although the human brain continues to mature well into early adulthood, the early years are a particularly sensitive period for “wiring” the developing brain, and compromised development during this time can increase the likelihood of aggression and violence. As illustrated, problematic development is more likely under conditions of extreme economic and social adversity. For some children, these early experiences can serve as stepping stones to an ill-fated course of habitual aggression and later violence, particularly if they are exacerbated by subsequent interactions at home, in school, and with peers. As Moffitt (1997) notes, “for poor urban children, the snowball of cumulative continuity is anticipated to begin rolling earlier, and it rolls faster downhill” (p. 150). The Early School Years (Ages 6 – 11) Many of the difficulties that begin before birth and through age 5 can be compounded by experiences when children enter school. Not only do educational opportunities and the structure of the school impact children’s development, but the influence of the peer group is heightened as children begin to learn rules for social engagement. Early school experiences can lead to academic problems, lack of engagement or “bonding” to school, and peer difficulties, all of which can increase risk for aggression and violence. In addition to the continued influence of communities and families, school entry opens (or shuts) new doors to academic and social development and can exacerbate the negative effects of earlier problems. For some children, new challenges may be particularly acute in light of earlier disadvantage. For example, children with early neuropsychological or IQ impairments associated with living in chronic poverty are likely to have difficulty learning basic academic skills. Schools in lowincome communities with limited resources may not provide remedial




Children learn both academic skills and social behavior at school. Although academic skills are linked to specific curriculum guidelines, a child’s social world is more likely to be regulated by the peer group. By the time children are 8 or 9 years old, peers demand more social competence and control of aggression. Aggressive children who are easily agitated and cannot solve social problems often are rejected by their peers, which can lead to increased aggression (Bierman, Smoot, & Aumiller, 1993). However, aggression does not always lead to peer rejection. In settings where aggression is normative and adaptive, aggression can lead to increased popularity and social status (Rodkin, Farmer, Pearl, & VanAcker, 2000). In economically disadvantaged communities with high rates of violence, being “tough” and ready to fight may be seen as an important quality from an early age, particularly for boys. During this time period, children become active agents in navigating and interpreting their social worlds. They acquire a cognitive understanding of the meaning of social interactions and develop characteristic styles of thinking that influence their behavior. In recent years, there has been an increased emphasis on the importance of the cognitive underpinnings of aggression and violence, particularly within psychology (Bandura, 1986; Crick & Dodge, 1994; Huesmann & Guerra, 1997). This line of research draws heavily from cognitive-information-processing theory, emphasizing both discrete social-information-processing skills as well as the “database” in memory that individuals develop over time. Much of the work in this area has emphasized sequential socialinformation-processing skills that involve encoding and interpretation of cues, response search, evaluation, decision, and action. Aggression has been associated with hostile attributional bias, attention to aggressive cues, generation of more aggressive solutions, and anticipation of positive outcomes (Crick & Dodge, 1994). However, a child’s choice of an appropriate response also hinges on what is encoded in memory as a normative standard for acceptable behavior. These standards develop from observation of one’s own behavior and the behavior of influential models as well as from direct instruction across contexts. As children get older, normative beliefs about aggression become increasingly stable and predictive of their own aggressive behavior (Huesmann & Guerra, 1997). In this regard, aggressive behavior in school, particularly when reinforced within this




developmental focus of the present chapter suggests, it is important to consider the cumulative effects of early experience on adolescent risk. In addition to a child’s being disengaged from school and forging an identity based on violence, the snowball of disadvantage can limit opportunities for career development and gainful employment, facilitate involvement in antisocial peer networks that require bravado rather than skill for personal and economic gain (particularly for males), and underscore the importance of the present over the future. It is here that the developmental tasks of adolescence intersect with age-specific and contextually bound constraints to solidify a path toward youth violence. Even for youth without an early history of aggression, the demands of adolescence coupled with limited opportunities for self-expression and advancement in poor neighborhoods may propel them to experiment with risky behaviors such as violence. Of course, adolescence is a time of increased engagement in risky behaviors for all youth. Recent advances in cognitive neuroscience have highlighted the role of brain maturation during adolescence in the etiology of these behaviors. From an evolutionary perspective, this is understandable. During this developmental stage, individuals must leave their families and seek out mates and opportunities for productive engagement — which require a certain degree of sensation seeking and risk taking. Distinct features of brain development that emerge in middle adolescence have been linked to increased reward seeking, specifically the rapid rise in dopaminergic activity around puberty. Simultaneously, self-regulation and coordination of affect and cognition are more problematic, linked to the slower maturation of the prefrontal cortex and its connection to other brain regions. At the same time, adolescence also is a time of considerable brain plasticity linked to experience (Steinberg, 2010). The developmental mandate to “leave the nest” also is accompanied by a psychological quest for a coherent identity. Identity includes a sense of uniqueness, a sense of continuity, and a sense of affiliation. Identity achievement has been heralded as a key task of adolescence, providing a link between past experience, present activities, and future goals (Erikson, 1968). Identity development is facilitated by advances in cognitive abilities during adolescence but is also impacted by available opportunities for productive engagement. In other words, who




their infants or leverage resources for academic enrichment are unlikely to have the time, skills, and fortitude to adequately supervise their children as they get older and become more independent. In addition to the effects of chronic poverty and short-term economic shocks, the effects of income inequality may be more pronounced for poor teenagers who grow up in a context of easily visible wealth. As their cognitive capacity for abstract thinking matures, inequalities become more salient, particularly for lower-social-status youth. Simultaneously, the need to stand out and be unique is amplified during this developmental stage, and possessions take on new meaning. In the United States and internationally, the lure of material wealth, strongly reflected in magazines, television, music, and the Internet, can further reinforce the sense of injustice experienced by many youth. Not only does inequality produce a type of cognitive disequilibrium, it has also been shown to increase chronic stress. Long-term exposure to this type of chronic stress shifts physiological priorities from those that maintain function to those that improve reaction times and provide quick energy for activity (Marmot & Wilkinson, 2006). As this illustrates, a cascade of events in poor, marginalized communities can create physiological and cognitive reactions that foster hypersensitivity to threat and danger from birth onward. It is easy to see how a spiral of cumulative disadvantage linked to socioeconomic factors can leave some youth unprepared for the developmental tasks of adolescence and increase risk for violence. Teenagers who have fallen behind academically are unlikely to have basic skills needed for advanced courses and to compete in a competitive job market. They are likely to be disconnected from school as a social institution, potentially limiting their social ties with conventional opportunities. Youth who do not find their niche in the school setting may seek out other like-minded youth and discover alternate opportunities in the underground or criminal economy. In neighborhoods where illegal opportunities are readily available and highly rewarding, the lure of easy money, camaraderie, and belonging may easily offset any negative consequences of this behavior. The focus on short-term rewards at the expense of long-term consequences characteristic of adolescent brain development is particularly acute for youth growing up in settings




during their lifetime, and most gang members only stay involved for about one year (Esbensen, Huizinga, & Weiher, 1993). What can be done to prevent youth violence under these difficult conditions? Absent the obvious shift to policies that engender a more equitable distribution of wealth (something which seems increasingly less likely in countries such as the United States), much still can be accomplished. This requires raising the floor of social inclusion, making critical services for healthy development universal, providing a more stable and equitable funding base for schools and training teachers to serve the needs of poor students, building opportunities and social connections in communities, teaching parents effective strategies for child rearing under adverse circumstances, and teaching children effective social-problem-solving skills from an early age. From a developmental perspective, early interventions that engage families and communities hold great promise for setting a strong foundation for child well-being and risk prevention. However, these efforts must be sustained, as the effects of early programs can easily wash out over time under harsh conditions. In addition to an early start, children also need boosters and “developmental cushions” along the way that protect them from the cumulative consequences of earlier mistakes. For children growing up under harsh conditions, relatively minor problems can easily initiate a downward spiral of person-environment interactions that can permanently foreclose future opportunities, absent their ability to bounce back from these events. Further, although there is general consensus that it is “never too early,” there is also a growing recognition that it is “never too late.” Indeed, remedial and second-chance programs for at-risk children and youth can work, even under conditions of extreme and chronic poverty and marginalization (Guerra, Williams, Walker, & Meeks-Gardner, 2010). Although much has been made of the importance of the early years for brain development, most studies point to “sensitive” rather than “critical” periods in the maturation of the brain (Knudsen, 2004), and the brain remains remarkably plastic into early adolescence (Steinberg, 2010). There are numerous examples of programs and policies that have improved children’s lives and have been effective in preventing aggression and violence. Many of these programs have been developed specifically




(Slavin, 1995). Large-scale projects such as the Comer project (Comer, 1998) have emphasized the importance of school reorganization in the most distressed urban settings in order to empower students, teachers, and parents to work together to maximize positive outcomes. A number of broader “social experiments” have demonstrated positive effects on child development and prevention of aggression and violence. Many of these community-based programs have involved relatively modest efforts to provide a safety net for poor families. For example, the New Hope project, operating in some of the poorest areas of Milwaukee, Wisconsin, in the 1990s, was designed to increase the benefits and to reduce the barriers of work by providing a range of available supports. These included an earnings supplement to raise income above the poverty line, subsidized child care, health insurance, temporary employment, and support from project staff. Among the most relevant findings, children from New Hope families, particularly boys, did better in school and displayed more prosocial and fewer aggressive behaviors than did comparison children (Duncan, Huston, & Weisner, 2006). Unfortunately, just as macroeconomic factors can lead to an array of community stressors and family disruptions that can impact child vulnerability and increase risk for childhood aggression and youth violence, resource constraints also can limit the availability of services to counteract or buffer the effects of these risk factors. Although programs such as home visitation and preschool enrichment historically have received high levels of governmental and private support, the infusion of funds into these programs may wane during difficult economic times. This may particularly impact early childhood programs during periods of heightened crime and violence, when benefits of these programs may not appear for many years. In spite of increasing support for the importance of early intervention, the politics of funding may preclude long-term investments in the face of problems requiring immediate solutions. On the other hand, skepticism about the effectiveness of remedial and second-chance programs can lead to an overall frustration with prevention programs and a default to punishment and suppression strategies. It is hoped that this chapter highlights the need to provide universal and targeted supports for the most at-risk children, supports that begin early in development and continue through adolescence.




Differences in explanatory variables. Journal of Contemporary Criminal Justice, 9, 94 – 116. Farrington, D. P. (1995). The development of offending and antisocial behavior from childhood: Key findings from the Cambridge study in delinquent development. Journal of Child Psychology and Psychiatry, 360, 929 – 964. Farrington, D. P. (2010). Developmental and life-course criminology: Key theoretical and empirical issues. Criminology, 41, 221 – 225. Fearon, R. P., Bakermans-Kranenburg, M. J., Van Ijzendoorn, M. H., Lapsley, A. M., & Roisman, G. I. (2010). The significance of insecure attachment and disorganization in the development of children’s externalizing behavior: A meta-analytic study. Child Development, 81, 435 – 456. Fleischner, J. E., & VanAcker, R. (1990). Monograph on critical issues in special education. Denton: University of North Texas Press. Gershoff, E. T. (2002). Corporal punishment by parents and associated child behaviors and experiences: A meta-analytic and theoretical review. Psychological Bulletin, 128, 539 – 579. Guerra, N. G. (1997). Intervening to prevent childhood aggression in the inner city. In J. McCord (Ed.), Violence and childhood in the inner city (pp. 256 – 312). New York: Cambridge University Press. Guerra, N. G., Huesmann, L. R., & Spindler, A. (2003). Community violence exposure, social cognition, and aggression among urban elementary-school children. Child Development, 74, 1507 – 1522. Guerra, N. G., Williams, K. R., Walker, I., & Meeks-Gardner, J. (2010). Building an ecology of peace in urban Jamaica. Washington, DC: World Bank. Hinshaw, S. P. (1992). Externalizing behavior problems and academic achievement in childhood and adolescence: Causal relationships and underlying mechanisms. Psychological Bulletin, 111, 127 – 155. Huesmann, L. R. (1998). The role of social information processing and cognitive schema in the acquisition and maintenance of habitual aggressive behavior. In R. Geen & E. Donnerstein (Eds.), Human aggression (pp. 73 – 109). San Diego, CA: Academic Press. Huesmann, L. R., & Guerra, N. G. (1997). Social norms and children’s aggressive behavior. Journal of Personality and Social Psychology, 72, 408 – 419. Jamaica Constabulatory Force. (2009). Island-wide violence trends. Ministry of National Security, Kingston, Jamaica. Jencks, C. (1993). Rethinking social policy: Race, poverty, and the underclass. New York: HarperCollins. Kandel, E., & Mednick, S. A. (1991). Perinatal complications predicts violent offending. Criminology, 29, 519 – 530. Knudsen, E. I. (2004). Sensitive periods in the development of the brain and behavior. Journal of Cognitive Neuroscience, 16, 1412 – 1425. Krivo, L. D., & Peterson, R. D. (1996). Extremely disadvantaged neighborhoods and urban crime. Social Forces, 75, 619 – 650.




Raudenbush, S. W. (1983). Utilizing controversy as a source of hypotheses for metaanalysis: The case of teacher expectancy’s effect on pupil IQ. Urban Education, 28, 114 – 131. Rodkin, P. C., Farmer, T. W., Pearl, R., & VanAcker, R. (2000). Heterogeneity of popular boys: Antisocial and prosocial configurations. Developmental Psychology, 36, 14 – 24. Schaeffer, C. M., & Borduin, C. M. (2005). Long-term follow-up to a randomized clinical trial of multisystemic therapy with serious and violent juvenile offenders. Journal of Consulting and Clinical Psychology, 63, 569 – 578. Slavin, R. E. (1995). Cooperative learning. Boston: Allyn & Bacon. Steinberg, L. (2010). A behavioral scientist looks at the science of adolescent brain development. Brain and Cognition, 72, 160 – 164. Straus, M., & Stewart, J. (1999). Corporal punishment by American parents: National data on prevalence, chronicity, severity, and duration, in relation to child and family characteristics. Clinical Child and Family Psychology Review, 2, 55 – 70. United Nations Office on Drug and Crime. (2009). International homicide statistics. Retrieved October 24, 2009, from Wilkinson, R. G., & Pickett, K. E. (2009). Income inequality and social dysfunction. American Sociological Review, 35, 493 – 511. Yoshikawa, H., Weisner, T. S., & Lowe, E. D. (2006). Making it work: Low-wage employment, family life, and child development. New York: Russell Sage Foundation.

10 Macroeconomic Factors and Inequities in Youth Violence The Cyclical Relationship between Community Conditions, Family Factors, and Youth Violence Jennifer L. Matjasko, Sarah Beth Barnett, and James A. Mercy

One of the central questions of this volume is whether there is a relationship between broader macroeconomic factors and youth violence.1 Macroeconomic fluctuations, or a deviation in the growth of gross domestic product (GDP) around a long-term trend, can cause systemic changes in institutions and macroeconomic conditions and at the same time cause stress or strain on local and regional economies, communities, families, and individuals. These economic fluctuations at the macro and local levels have differential effects on individuals that may depend on their community conditions as well as their families’ socioeconomic status (Moen, 1979). Research on the risk factors for youth violence involvement, for example, points to the importance of the community and family contexts in predicting long-term involvement in violence (e.g., Sampson, Wilson, & Petersilia 1995). Furthermore, com278



munities that are differentially affected by economic fluctuations often have higher rates of crime and youth violence involvement (Bursik & Grasmick, 1993). Oftentimes, these communities have persistently high rates of violence that may be driven by long-term inequities in youth violence that can occur over many years and many generations (Land, McCall, & Cohen, 1990). Based on existing theory and research, we posit theory that specifies a cyclical relationship between community economic factors and youth violence that may explain why racial inequities in youth violence exist and persist in the United States. This theory is described in this chapter. Racial inequities in rates of youth violence, especially in its more severe forms (e.g., homicide), have existed for as long as data on this phenomenon have been collected. Because of the high correlation between race and socioeconomic status, family and community economic conditions may partially explain these inequities. There are stark racial differences in youth arrests, differences that are persistent over time (Elliott, 2009). In the National Youth Survey, the African American – Caucasian race gap in those who report serious violence is the greatest between the ages of 14 and 17 (Elliott, 2009). The rate of African American involvement with serious violence is 50% higher than the rate of serious violence involvement among Caucasian youth (U.S. Department of Health and Human Services, 2001). In addition, in 1998, there were approximately five African American youths arrested for homicide for every Caucasian youth. This racial gap is much greater for arrests compared to self-reported violence (U.S. Department of Health and Human Services, 2001). Nevertheless, the gap in arrests has persisted over time. An important question is why these inequities have persisted. On the surface, one may surmise that since they have persisted over a wide range of macroeconomic fluctuations, these fluctuations may have little to do with them. However, there may be a dynamic relationship between more localized community and family economic conditions and youth violence, in which both types of economic conditions and youth violence replicate one another. One explanation for the racial inequities in youth violence is neighborhood-concentrated disadvantage (Pratt & Cullen, 2005). African Americans are more likely to live in disadvantaged neighborhoods due to community conditions such as




models that link macroeconomic fluctuations and community economic conditions with child development; (c) describe the impact of stress on child development; (d) describe how the social environment is linked to developmental pathways of youth violence; (e) show how youth violence affects the economic conditions of a community; (f) posit a cyclical model of economic factors and youth violence; and (g) end with some modeling, policy, and program implications of this cyclical model. The purpose of this chapter is to review existing theory and research in order to construct a plausible theoretical model that helps to explain the persistence of racial and socioeconomic inequities in youth violence by linking macroeconomic factors with community and family factors known to be associated with youth violence.

Defining Macroeconomic Fluctuations and Conditions A macroeconomic fluctuation is a deviation in the growth of gross domestic product around a long-term trend. There are several types of macroeconomic fluctuations, including business cycles, shocks, bubbles and crises. Although the main outcomes of fluctuations are changes in GDP, the way they impact people’s lives and their root causes are different; therefore, policies aimed at mitigating the effects of fluctuations on the propensity of youth to commit violence must be catered to the type and determinants of fluctuations. For that reason, it is important to define and draw the distinction between differing types of macroeconomic fluctuations. The business cycle is a source of common macroeconomic fluctuations, because the economy is continually in one of four phases of GDP growth: expansion, peak, recession, and trough. Expansions are periods of growth in real (adjusted for inflation) GDP, and recessions are defined as two consecutive quarters of negative GDP growth (O’Sullivan & Sheffrin, 2003). Shocks are events that are unexpected or unpredictable, and in economic modeling terms, they are exogenous, or external to the economic system. Shocks subsequently push the economy into disequilibrium and result in changes in macroeconomic factors and outcomes. Shocks happen frequently, though the severity of their effects on markets and the economy vary widely. A prime example is the terrorist attacks that took place on 9/11, which were unpredictable and had numerous effects on the economy,




part of the San Francisco Bay Area, is home to some of the largest technology companies and was affected by the many dot-com companies that went under in the early 2000s. We refer to the economic factors and environment that individuals and families interact with at the local level as community economic conditions. We turn our attention to the impact of macroeconomic fluctuations, local economy fluctuations, and community economic conditions on youth violence. The primary mechanism through which macroeconomic fluctuations may influence youth violence involvement may be through community and family economic conditions. Community economic conditions include local job opportunities, local sectors and industries, credit markets, local policies related to small businesses, taxes, and community resources such as nonprofit organizations. Family economic conditions include family income and benefits, assets (home ownership, retirement or college savings), and access to credit (borrowing and indebtedness). Families and youth in lower socioeconomic strata may be more severely disrupted by macroeconomic fluctuations than are those in higher strata because they are more likely to lack the familial and community resources to buffer the impact of such fluctuations. In other words, macroeconomic fluctuations may contribute to community and familial environments that are disadvantaged and strained. Community and family strain are related to stressors such as family disruption and harsh parenting (Webster-Stratton, 1990). When such stressors on children are chronic, uncontrollable, and experienced without the support of caring adults, they can influence brain development in ways that may contribute to risk for youth violence (National Scientific Council on the Developing Child, 2005). In a very real sense, macroeconomic fluctuations and community and family economic conditions, therefore, may alter children’s developmental trajectories and result in a cycle of disadvantage. This cycle of disadvantage may also be associated with a cycle of relatively higher rates of youth violence in disadvantaged communities, compared to the rates in communities with greater economic resources. In the next section, we present some theoretical models that describe the relationship between macroeconomic factors, communities, families, and child development.




Fig. 10.1. Above, the family stress model Fig. 10.2. Left, the family investment model Fig. 10.3. Below, general strain theory

(Bronfenbrenner, 1979). The social ecological model (Bronfenbrenner, 1979) is a model wherein the individual is nested within relationships, the community, and the larger society. In this case, the macroeconomic fluctuations are part of the “outer levels” of the social ecology, in that these fluctuations occur at city, state, or national levels.




investment model may be more relevant for families that face more persistently poor family economic conditions, such as poverty. Evidence for this model comes from studies that document that family income during childhood and adolescence is positively correlated with academic, financial, and occupational success in adulthood (Bradley & Corwyn, 2002; Corcoran & Adams, 1997; Mayer, 1997; Teachman, Paasch, Day, & Carver, 1997). Perhaps the strongest evidence comes from researchers who used several waves of the National Longitudinal Survey of Youth (Bureau of Labor Statistics, 2002) and found significant differences related to family investments in enriching home environments among families living above versus below the poverty line (Bradley, Corwyn, McAdoo, & Garcia Coll, 2001). These investments included both social and physical resources such as the quality of parent-child relationships and the provision of cognitively enriching activities. Income and investment are expected to translate to developmental success through these family processes. Researchers have found an association between child cognitive development at ages 3 and 5 and family income, controlling for parental education and intelligence, establishing a relationship between income and child functioning over and above these parental characteristics (Linver et al., 2002). Both the family stress and family investment models contain some assumptions about the nature of family economic conditions. The family stress model frames economic stressors as decreases in income, presumably from a macroeconomic fluctuation since the model was developed based on a downturn in the agricultural economy in the 1980s (Conger et al., 1992). The family investment model usually defines the family economic condition as family disadvantage or poverty, a term often used to describe a state that is persistent. When families are in poverty, they tend to face constant economic stress and financial strain. Macroeconomic fluctuations such as recessions that result in unemployment may place further strain on families who are already unemployed and living below the poverty line. The family investment model posits that families exposed to long-term economic disadvantage already face chronic economic stressors, and macroeconomic fluctuations may result in additional strain on them. These two models appear to be pointing to two distinct experiences when it comes to macroeconomic factors — one subset of families may be influenced by macroeconomic




of the economic stressor (i.e., is it a short- or a long-term event?), and (3) the timing of the economic stressor (i.e., at what point in development does it occur?). With short-term macroeconomic fluctuations, all three of these considerations are sources of variability that can elucidate the relationship between macroeconomic fluctuations, community and family economic conditions, and youth violence. On the other hand, the concept of long-term or chronic economic conditions assumes that there is persistence to the stressor and that it occurs throughout development. However, the nature of that stressor can vary. In addition, there can be variation over time in the nature of this stressor, so it is also important to tie the nature of the stressor to developmental periods over the course of childhood and adolescence. When economic conditions are severe, occur early in development, and are persistent over time, it is likely that the conditions will negatively influence development and increase the likelihood of youth violence involvement. In the following section, we describe the developmental literature on youth violence involvement.

Social Environment, Child Development, and Youth Violence Macroeconomic fluctuations can contribute to youth violence in at least two key ways. First, they can impact families and increase children’s exposure to key risk factors for youth violence, such as harsh parenting. Second, they can contribute to community conditions that lead to greater social disorganization and youth violence in communities by, for example, creating communities with larger numbers of delinquent and antisocial youth. The developmental literature on youth violence involvement has identified developmental-stage-specific risk factors for youth violence as well as various developmental trajectories of youth violence. Most of this work is based on longitudinal studies of youth violence involvement. In this volume, Guerra (chapter 9) provides an excellent summary of the developmental-stage-specific risk factors for youth violence for the prenatal period (through age 5), the early school years (ages 6 – 11), and adolescence (ages 12 – 18). We present a summary of these developmental stage risk factors in table 10.1. Guerra emphasizes the cumulative nature of stress throughout development and that macroeconomic stressors are typically filtered through families




or otherwise achieve higher socioeconomic status (Guerra, chapter 9 in this volume). As a result, communities also tend to be locked in a cycle of economic disadvantage.

A Cyclical Model of Economic Factors and Youth Violence Based on the theory and research just presented, we posit a cyclical, or recursive, relationship between community and family economic factors and youth violence. In essence, we are suggesting that a dynamic system exists that maintains and sustains a relationship between community and family economic factors and youth violence and that resists disruption. This theory seeks to explain why racial and socioeconomic inequities in youth violence have persisted over a long period of time in the United States. Figure 10.4 depicts this model. The main elements of the model are (1) economic fluctuations are filtered through community and family economic conditions, and this predicts parenting, strain, and school/peer risk factors; (2) the extent to which economic fluctuation influences parenting, strain, and school/ peer risk factors depends on the extant community and family economic conditions (represented by the dashed line in figure 10.4); (3) the experience of prolonged and substantial economic strain within communities and families disrupts healthy patterns of child development by contributing to family instability and undermining a family’s ability to

Fig. 10.4. Community and family economic conditions and youth violence: a cyclical model




assumption; if there is a hypothesized lag between a larger macroeconomic fluctuation and youth violence, the model can be estimated with a lagged measure of macroeconomic shocks. The community and family economic conditions variables can be added in a separate model to test their strength as a mediator and a moderator. In other words, community and family economic conditions may mediate the relationship between macroeconomic fluctuations and parenting, strain, and school/peer risk factors, which are related to individual risk factors and youth violence. In order to accurately test mediation, the other mediating variables — parenting/parental investments, strain, and school/peer risk factors — can also be added in separate models. Furthermore, the extent to which macroeconomic fluctuations affect parenting, strain, and school/peer risk factors depends on the community and family economic conditions at the time the fluctuation occurs. Thus, the interaction between macroeconomic fluctuations and community/family economic conditions in predicting those mediating variables would also be included in the model. Finally, individual risk factors can then be added to the model. Evidence of mediation can be shown by significant paths to and from the mediating variables, as well as a reduction in the direct effects of the exogenous variable on the outcome. Finally, a structural equation model can be used to test whether youth violence at time t is significantly related to community and family economic conditions at time t + 1.

Prevention and Policy Implications In this chapter, we made the distinction between two types of economic factors — ones that are a result of short-term macroeconomic fluctuations and long-term economic conditions that are manifested at the community and family levels in the form of community and family economic conditions. We presented the family stress and investment models as well as general strain theory — three models that have been used to explain how macroeconomic fluctuations within local economies influence communities and families, which, in turn, are thought to influence youth violence. We integrated these three models and proposed a cyclical model that specifies the relationship between macroeconomic fluctuations, community and family economic conditions,




poverty-reduction policies transfer income and wealth from some individuals to others, often through taxing individuals with higher incomes and distributing the revenue to low-income individuals (Alesina & La Ferrara, 2005). Another potentially viable strategy to alleviate disadvantaged community and family economic conditions and perhaps vulnerability to macroeconomic fluctuations is the establishment of business improvement districts (BIDs) or other strategies to create a more viable community economic environment in distressed communities (Levy, 2001). Existing research is inconclusive about the effectiveness of BIDs in reducing youth violence. It is clear that BIDs alone cannot create and sustain neighborhood change, but some research has suggested that many BIDs have invested in private security measures (MacDonald, Golinelli, Stokes, & Bluthenthal, 2010). In addition, BIDs have been found to have significant positive effects on reducing rates of crime and, in particular, robberies (MacDonald et al., 2009). Other ways to reduce long-term community and family economic conditions include job training, living-wage policies, and policies that reduce discriminatory lending practices (e.g., higher-interest-rate loans to minorities) so that low-income individuals can acquire higher earnings and property, respectively. It is also important to evaluate whether these economic policy approaches are effective at shortening the duration of macroeconomic fluctuations and reducing their impact on community and family economic conditions and youth violence. Developmental Considerations for Youth Violence Prevention Despite policies aimed to alleviate macroeconomic fluctuations and long-term community and family economic conditions, some income inequity within the United States is probably inevitable. Therefore, it is also important to bolster families’ ability to cope with disadvantaged economic conditions in order to reduce the likelihood that youth will become involved with violence. First, for macroeconomic fluctuations, it might be necessary to focus on families that may be economically vulnerable or at the margin, since those are the families that are likely to feel the impact of such fluctuations (Stiglitz, 1999). For both economically vulnerable families and those that have experienced long-term poverty and inequality, it is important to provide supports that bolster family




available to youth and their families because of budget cuts. When this combination of events occurs, vulnerable families have fewer resources and sources of support to help them cope with macroeconomic fluctuations and disadvantaged economic conditions. Decisions to cut school and local budgets in such ways may thus have unintended consequences on youth violence, making the availability of prevention programs at multiple levels of the social ecology even more vital (for a review, see Limbos et al., 2007). In sum, a multifaceted effort at multiple levels is necessary in order to break the cycle of disadvantage and to reduce inequities in youth violence. Notes 1. Disclaimer: The findings and conclusions of this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

References Agnew, R. (1999). A general strain theory of community differences in crime rates. Journal of Research on Crime and Delinquency, 36, 123 – 155. Alesina, A., & La Ferrara, E. (2005). Preferences for redistribution in the land of opportunities. Journal of Public Economics, 89, 897 – 931. Ames, B., Brown, W., Devarajan, S., & Izquierdo, A. (2001). Macroeconomic policy and poverty reduction. Washington, DC: International Monetary Fund. Becker, G., & Thomes, N. (1986). Human capital and the rise and fall of families. Journal of Labor Economics, 4, S1 – S139. Benson, M. L., & Fox, G. L. (2004). When violence hits home: How economics and neighborhood play a role (Research in Brief). Rockville, MD: National Institute of Justice. Bradley, R. H., & Corwyn, R. F. (2002). Socioeconomic status and child development. Annual Review of Psychology, 53, 371 – 399. Bradley, R. H., Corwyn, R. F., McAdoo, H. P., & Garcia Coll, C. (2001). The home environments of children in the United States: Part I, Variations by age, ethnicity, and poverty status. Child Development, 72, 1844 – 1867. Bronfenbrenner, U. (1979). The ecology of human development. Cambridge: Harvard University Press. Bureau of Economic Analysis. (2010). Gross domestic product: Percent change from preceding period. Bureau of Labor Statistics. (2002). National Longitudinal Survey of Youth 1979 cohort, 1979 – 2002 (rounds 1 – 20). Produced and distributed for the U.S. Department of




Haveman, R. B., & Wolfe, B. (1994). Succeeding generations: On the effects of investments in children. New York: Russell Sage Foundation. Jargowsky, P. A. (1997). Poverty and place: Ghettos, barrios, and the American city. New York: Russell Sage Foundation. Johnson, N., Williams, E., & Oliff, P. (2010). Governors’ new budgets indicate loss of many jobs if federal aid expires. Washington, DC: Center on Budget and Policy Priorities. Knitzer, J. (2000a). Early childhood mental health services: A policy and systems development perspective. In J. P. Shonkoff & S. J. Meisels (Eds.), Handbook of early childhood intervention (2nd ed., pp. 416 – 438). New York: Cambridge University Press. Knitzer, J. (2000b). Promoting resilience: Helping young children and families affected by substance abuse, domestic violence, and depression in the context of welfare reform (Children and Welfare Reform Issue Brief No. 8). New York: National Center for Children in Poverty. Land, K. C., McCall, P. L., & Cohen, L. E. (1990). Structural covariates of homicide rates: Are there any invariances across time and social space? American Journal of Sociology, 95, 922 – 963. Levy, P. (2001). Paying for the public life. Economic Development Quarterly, 15, 124 – 131. Limbos, M. A., Chan, L. S., Warf, C., Schneir, A., Iverson, E., Shekelle, P., & Kipke, M. D. (2007). Effectiveness of interventions to prevent youth violence: A systematic review. American Journal of Preventive Medicine, 33, 65 – 74. Linver, M. R., Brooks-Gunn, J., & Kohen, D. (2002). Family processes as pathways from income to young children’s development. Developmental Psychology, 38, 719 – 734. Luthar, S., Cicchetti, D., & Becker, B. (2000). The construct of resilience: A critical evaluation and guidelines for future work. Child Development, 71, 543 – 562. MacDonald, J., Blumenthal, R. N., Golinelli, D., Kofner, A., Stokes, R. J., Sehgal, A., Fain, T., & Beletsky, L. (2009). Neighborhood effects on crime and youth violence: The role of business improvement districts in Los Angeles. Santa Monica, CA: RAND. MacDonald, J., Golinelli, D., Stokes, R. J., & Bluthenthal, R. (2010). The effect of business improvement districts on the incidence of violent crimes. Injury Prevention. Mayer, S. (1997). What money can’t buy: Family income and children’s life chances. Cambridge: Harvard University Press. McLoyd, V. C., Jayaratne, T. E., Ceballo, R., & Borquez, J. (1994). Unemployment and work interruption among African American single mothers: Effect on parenting and adolescent socioemotional functioning. Child Development, 65, 562 – 589. Mistry, R. S., Vandewater, E. A., Huston, A. C., & McLoyd, V. C. (2002). Economic well-being and children’s social adjustment. Child Development, 73, 935 – 951. Moen, P. (1979). Family impacts of the 1975 recession: Duration of unemployment. Journal of Marriage and the Family, 41, 561 – 572. National Scientific Council on the Developing Child. (2005). Excessive stress disrupts the architecture of the developing brain (Working Paper #3). http://www




Surgeon General. Rockville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Injury Prevention and Control; Substance Abuse and Mental Health Services Administration, Center for Mental Health Services; and National Institutes of Health, National Institute of Mental Health. Webster-Stratton, C. (1990). Stress: A potential disruptor of parent perceptions and family interactions. Journal of Clinical Child and Adolescent Psychology, 19, 302 – 312. Werner, E. E. (2000). Protective factors and individual resilience. In J. P. Shonkoff & S. J. Meisels (Eds.), Handbook of early childhood interventions (2nd ed., pp. 115 – 132). New York: Cambridge University Press. Yeung, W. J., Linver, M. R., & Brooks-Gunn, J. (2002). How money matters for young children’s development: Parental investment and family process. Child Development, 73, 1861 – 1879.

This page intentionally left blank

11 Economic Opportunity and Youth Violence Conclusions and Implications for Future Research Curtis S. Florence and Sarah Beth Barnett

In this volume, the authors have sought to identify macroeconomic factors that are associated with youth violence (defined as violence where the victim or perpetrator is 10 to 24 years old) and to explore aspects of the relationship between economic conditions and violence that can be useful in developing interventions to prevent violence.1 In conducting the project, we departed from the standard textbook definition of macroeconomics — economic factors at the national, regional, or international level. Instead, we include economic factors across all levels of aggregation, from the national to the family level. Our reason for doing this was to examine economic factors that can potentially affect individual behavior via their impact on neighborhoods, schools, street markets, and family functioning. Thus, we examine factors that are likely to influence violence at the level of society and community — the higher >>





effective programs that prevent violence through the improvement of economic conditions, this project can provide valuable information for understanding the evaluation of current programs and the development of future programs. We will review and synthesize the results of the chapters in this volume by viewing them in the context of the three issues raised in the preceding paragraph. While it has not been the purpose of this project to evaluate specific programs or policies to prevent violence by improving economic conditions, the findings reviewed here can help lay the groundwork for developing and evaluating those interventions in the future.

The Connection between Economics and Violence There has been a large literature produced that examines the relationship between economic factors and crime. This literature has examined a wide variety of criminal outcomes and economic variables but generally attempts to establish whether crime is “countercyclical” — that is, whether crime goes up when economic activity declines and decreases when economic activity increases. In general, this literature focuses on the temporal associations between economic factors and crime in order to examine the effects of economic conditions net of the impact of confounding variables that may be associated with the levels of economic variables and crime. One of the seminal studies in this area was Cook and Zarkin (1985), which studied the relationship between recessions and property crime (auto theft and burglary) and violent crime (robbery and homicide). This study used national crime statistics published by the FBI and found a significant countercyclical relationship between both burglary and robbery and recessions but not a significant relationship with homicide. In this study, each recession since the 1940s was treated as its own experiment, and the relationship between the onset of the recession and the growth rates of crimes was tested. Additional tests were conducted with regression models that estimated crime as a function of either the unemployment rate or the employment ratio (the percentage of the total population employed), and the results were very similar to the associations of crime with recessions. The Cook and Zarkin studies focus on unemployment as an indicator of overall macroeconomic conditions; this work has subsequently been reexamined by




victims and perpetrators below age 25, who are at particularly high risk for violence. This raises an interesting possibility. The lack of a significant relationship between economic conditions and homicide found in Cook and Zarkin (1985) (as well as more recent studies such as Arvanites & Defina, 2006) could be driven by a combination of differential impacts of changes in employment status and income on violence, a combination that depends on the age of the victims and perpetrators. Therefore, it is a possibility that changes in income have a different impact on youth homicide than on homicide in other age ranges. Given the interesting possibility raised by these results, it is fortuitous that these issues are examined in two other studies in this volume. Baumer, Rosenfeld, and Wolff extend the research on economic factors and crime along several dimensions. A particularly interesting aspect of their study is that they use data that is disaggregated from the national level, by collecting data on property crime and homicide from 82 metropolitan areas around the country. A notable strength of their study is that they include data for the most recent economic downturn. The authors control simultaneously for several economic conditions, including the unemployment rate, the mean wage rate, real GDP, and an index of consumer sentiment. They also measure the homicide rate for perpetrators under 18, 18 to 25, and over 25, so outcomes are examined for youth homicide as well as overall homicide. A key advantage of this approach is that the authors examine violence outcomes at a lower level of aggregation, so the economic factors examined should not suffer the same amount of collinearity as the national data examined by Bushway, Cook, and Phillips. The more recent data allows for examination of the most recent recession. While a key focus of the chapter is the moderating effects of inflation and policy variables on the effects of economic conditions, this chapter will focus on the linear effects found for the various economic factors. Baumer, Rosenfeld, and Wolff ’s results show that higher real wages are associated with lower homicide rates overall and for youth under 18. However, higher real wages are also positively associated with higher rates of homicide for youth 18 to 24. Higher real GDP is associated with lower homicide rates overall and for youth 18 to 24 and adults over 25. Higher consumer sentiment (a measure of how positively the public




By comparing the results across these three studies, some common themes are apparent. First, examining the relationship between macroeconomic conditions and violence is much more complicated than asking whether a bad economy results in more violence. Previously published studies of this relationship have generally examined data collected at a high level of aggregation (e.g., national or state) and have not examined youth violence specifically. However, when the data begin to be disaggregated from the national level and measures of income are included among the economic factors examined, a clearer picture of the relationship between macroeconomic factors and youth violence begins to emerge. Measures of income are more likely to be associated with violence than are measures of employment. In some cases, these measures of income are more closely associated with violence among youth than they are with adults. These studies are also consistent with results that have examined the relationship between wages and violence (Gould, Weinberg, & Mustard, 2002). The strikingly consistent results across the three studies points to a future area of research that could be useful for violence prevention: why does there appear to be a specific relationship between income and violence that is not seen between violence and other economic variables? If there is a direct connection between income and violence, interventions to increase income and wages may be effective in preventing violence.

Neighborhoods, Schools, and Street Markets While the studies in this volume have documented the connection between certain macroeconomic conditions and violence, this raises the question of how economic factors such as income at a national, state, or local level relate to individual violent behaviors. We will now review the studies in this volume that examine the effect of economic conditions through the mediating domains of neighborhoods, schools, and street markets to examine factors that are still macroeconomic in the sense that they measure conditions at a higher level than the individual or family. These factors are still broadly experienced by a large group of people and, therefore, fall under the societal and community levels of our explanatory framework.




that have moved families from public housing to areas of less concentrated poverty. The Moving to Opportunity (MTO) demonstration project provided housing vouchers to low-income families living in areas of concentrated poverty to allow them to move to areas of lessconcentrated poverty. Studies of the effects of this program found a significant reduction of violence by girls, but after initially positive results for boys, the effects on violence became negative (although nonsignificant; Kling, Ludwig, & Katz, 2005). The Fang et al. results suggest that these mixed results could arise because the improvement in neighborhood conditions induced by MTO may not have been sufficient to reduce violent behavior on the part of boys. This is consistent with the conclusions of the final report on the MTO evaluation, which stated that “a more comprehensive approach is needed [than housing vouchers] to reverse the negative consequences of living in neighborhoods with heavily concentrated poverty” (U.S. Department of Housing and Urban Development, 2011, p. 7). The results of Fang et al.’s study also have implications for other interventions that attempt to improve neighborhood conditions to reduce crime and violence, such as business improvement districts (BIDs). BIDs allow local business owners to vote to collect from local businesses fees that are used to invest in neighborhood improvements in the district where they are collected (examples of BID investments in Los Angeles include security patrols, maintenance of sidewalks, graffiti removal, marketing of local businesses, parking and transportation management, and capital improvements, such as improved lighting).3 This strategy has been shown to lead to a significant reduction in assaults following implementation of BIDs in Los Angeles districts (Cook & MacDonald, 2011). As this program is adopted in other locations, it will be important to view the Los Angeles experience in the context of the Fang et al.’s results. Did the BID program allow highly disadvantaged neighborhoods to improve substantially in ways that led to reduced violence? Was the program implemented in districts that were on the verge of moving into the middle ranges of disadvantage? Addressing these questions will be important to understand how the BID program has led to the improvements demonstrated. An additional study included in this volume, by Crutchfield and Wadsworth, also examines the effects of neighborhood disadvantage




West’s result is confirmed by further research, it will add an important component to the evaluation of crime control policies. Finally, the relationship between neighborhood conditions and street markets is potentially an important pathway for macroeconomic conditions to influence violent behavior. For example, some researchers have linked the dramatic decline in homicide in the 1990s to the waning of the crack-cocaine market, which was characterized by a high level of violence (Blumstein & Wallman, 2000). Edberg and Bourgois present the results of a qualitative study that examines the relationship between the identity formation of adolescents and the violent imagery surrounding narcotraffickers in the U.S.-Mexico border region. It is clear that the positive portrayal of participants in the drug trade who are engaged in violent behavior strongly influences the self-image of the participants in the study. However, it remains an open question how much the presence of drug trafficking affects adolescent development and violence outcomes when other factors are held constant. In addition, it is not clear that this phenomenon experienced in the U.S.-Mexico border region generalizes to the rest of the United States. Given the illegal nature of most street markets, it remains a very difficult area to address with the type of quantitative analysis that is necessary to establish the net effect of street market activity on violence. The data-collection challenges are daunting and will likely involve substantial resources to accomplish. For example, Ihlanfeldt (2007) examined the effect of the distance to available jobs on illegal drug sales in Atlanta, Georgia, neighborhoods. However, violence associated with the drug trade was not specifically examined. This study involved geocoding thousands of crime records from multiple law enforcement agencies compiled over a two-county area for four years. The author states that the difficulty in collecting data precluded examining additional years of data. This shortcoming will need to be addressed if the effect of street markets on violent behavior is to be evaluated and effective interventions are to be developed. The work in this volume illustrates that neighborhood, school, and street markets are important mediators that filter broader macroeconomic conditions down to the individual behavior. While the causal links among these three conditions are complex and not easily teased out in a research setting, understanding the




the American Academy of Pediatrics published a policy statement in which it supports leveraging science and early childhood policies and services to reduce the precipitants of toxic stress and to mitigate their impact (Garner et al., 2012). Since both chronic poverty and economic shocks can contribute to developmental risks and toxic stress, it is important to study the impact of economic fluctuations on youth violence rates many years after the economic change. Guerra states, “most studies examine changes in violence rates within the defined period during or immediately following economic downturns. Yet many of the impacts on youth violence may affect children even before they are born, translating into changes in youth violence rates a decade or more in the future” (this volume, p. 258). Despite the large body of evidence that demonstrates the impact of early life stress on child development, there have been very few studies that have examined the long-term impacts of severe economic downturns on child health and behavior. One exception is Elder (1999), which examined the impact of the Great Depression on children’s life course. Elder’s study found that the Depression negatively affected working-class children more than middle-income children. Economic downturns may disproportionately impact families that are living in poverty or are financially unstable. Future studies must document both the impact of economic conditions on families and their long-term impacts on children and youth violence rates. Matjasko et al. discuss existing child development theory and propose a framework for explaining the interworking of macro, community, and family economic factors and their impact on family functioning. An existing model, the family stress model (Conger et al., 1992), posits that macroeconomic factors affect family economic factors that then alter family stress, parental emotion and behavior, and parenting practices. The family economic factors that could be affected include debt accumulation, asset loss, damage to property (e.g., the inability to pay for car repairs), housing transition (e.g., foreclosure or eviction), loss of medical coverage, food insecurity, and so on. These types of strain affect both chronically poor families as well as families that do not have the resources to withstand disruptions, including families that are near poverty. The family investment model (Becker & Thomes, 1986) underscores the relative differences in families’ ability to promote




by Blueprints for Violence Prevention (Mihalic et al., 2001). The program sends trained registered nurses to visit first-time mothers on an ongoing basis while they are pregnant through the second year of the child’s life. The nurses promote positive health-related behaviors during pregnancy and the early years of the child’s life, competent care of the children, and maternal personal development. The nurses help women envision a future consistent with their values and aspirations (Olds, 2002). These aspects of the program reduce negative childhood experiences, promote children’s resilience and development, and support positive family functioning in the face of economic strain. Since most participants are low income, these skills may be especially salient in the setting of economic instability or poverty. Furthermore, nurses also help mothers with family planning, connect them to government services, and help women continue their education and find employment. NFP acknowledges the gravity of economic factors and integrates economic factors into the program. Economic self-sufficiency for the family is one of NFP’s three stated goals (NFP, 2012). The program has demonstrated some long-term effects on youthviolence-related behaviors. In a 15-year follow-up of the first randomized control trial in Elmira, New York, children in the experimental group reported fewer arrests and convictions (Olds et al., 1998). However, in a 19-year follow-up, it appears that this effect only continued for girls (Eckenrode et al., 2010). Some of the long-term benefits of NFP may result from the way in which the program incorporates economic factors. For example, mothers who participate in NFP are more likely to be employed over time (Olds et al., 1988; Kitzman et al., 1997; Olds, 2009) and have longer intervals between first and second children due to family planning (Olds et al., 1988; Kitzman et al., 2000; Olds et al., 2004). Economic evaluations of both the Elmira and Memphis pilots show that the program is cost saving to society. Most of the benefits of the program come from mothers in the treatment groups using less government services including AFDC or TANF, food stamps, and Medicaid (Karoly et al., 1998; Olds et al., 2010). In addition to early home-visitation programs, there is also evidence that early childhood education programs can reduce violence later in life. Rigorous evaluations of high-quality, center-based preschool programs for three- and four-year-old low-income children that include an




economic conditions and violence all identified significant relationships. The nature of these relationships, however, is not consistent across all economic conditions or all age groups. These findings point to a strong countercyclical relationship between income, wages, and youth violence. None of the studies identified a significant relationship between unemployment and youth violence, although this has been the primary focus of much of the prior research on economic conditions and crime. Given the relative void in the research literature on the relationships between income and youth violence, this is an area where future research can be helpful in describing the nature of this relationship can and provide information that can be useful in the development of preventive interventions that focus on this economic factor. It would be useful to know whose income is important. Is it family income? Income earned by youth and young adults? Addressing these questions will assist in designing programs and policies that can reduce violence by increasing income. Second, if economic factors are associated with violence, policies, programs, or strategies developed to address these conditions must be able to identify the appropriate economic characteristics and be effective in improving them. While the studies in this volume did not directly examine the effectiveness of particular preventive intervention in improving economic conditions, they did examine several mediating domains that are potential pathways between higher-level macroeconomic conditions and youth violence outcomes. In particular, the nature of the relationship between neighborhood economic and societal conditions and youth violence was shown to be complex but extremely vital in understanding the relationships between macroeconomics and violence. Neighborhood disadvantage was shown to be associated with violent behavior by youth, but the relationship is not the same for all neighborhoods. In order to decrease violence in severely disadvantaged neighborhoods, it is likely that the necessary improvements in economic conditions would need to be large. There is also a complex relationship between neighborhood conditions and grades in school. While higher grades in school are generally associated with less delinquency by youth, the results presented in this volume suggest that for youth in severely disadvantaged neighborhoods, this is not the case. It will be important for future research




effect may be due to a substantial improvement in family economic conditions. This is an important preventive pathway for future program evaluations to examine. In summary, the research presented here provides strong evidence of a relationship between certain macroeconomic conditions and youth violence. The associations between economic conditions and violence are likely mediated by several domains and in complex ways. However, violence is preventable, and the prevention of violence through improving economic conditions is a promising pathway. In the future, it will be important to better understand the short- and long-term relationships between macroeconomics and violence and to evaluate prevention programs for their effect on economic outcomes. Notes 1. Disclaimer: The findings and conclusions of this chapter are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention. 2. Guide to Community Preventive Services, index.html. 3. Office of Justice Programs, “Program Profile: Business Improvement Districts (BIDs), Los Angeles (Calif.),” .aspx?ID=67.

References Arvanites, T. M. & Defina, R. H. (2006). Business cycles and street crime. Criminology, 44(1), 139 – 164. Becker, G., & Thomes, N. (1986). Human capital and the rise and fall of families. Journal of Labor Economics, 4, S1 – S139. Blumstein, A., & Wallman, J. (2000). The recent rise and fall of American violence. In A. Blumstein & J. Wallman (Eds.), Crime drop in America (pp. 1 – 12). Cambridge: Cambridge University Press. Conger, R. D., Conger, K. J., Elder, G. H., Lorenz, F. O., Simons, R. L., & Whitbeck, L. B. (1992). A family process model of economic hardship and adjustment of early adolescent boys. Child Development, 63, 526 – 541. Cook, P. J. & MacDonald, J. (2011). Public safety through private action: An economic assessment of BIDs. Economic Journal, 121(552), 445 – 462. Cook, P. J., & Zarkin, G. A. (1985). Crime and the business-cycle. Journal of Legal Studies, 14(1), 115 – 128. Dahlberg, L. L., & Krug, E. G. (2002). Violence — a global public health problem. In E.




Olds, D. L., Henderson, C. R., Cole, R., Eckenrode, J., Kitzman, H., Luckey, D., et al. (1998). Long-term effects of nurse home visitation on children’s criminal and antisocial behavior. JAMA, 280(14), 1238 – 1244. Olds, D. L., Henderson, C. R., Tatelbaum, R., & Chamberlin, R. (1988). Improving the life-course development of socially disadvantaged mothers: A randomized trial of nurse home visitation. American Journal of Public Health, 78(11), 1436 – 1445. Olds, D. L., Kitzman, H., Cole, R., Hanks, C. A., Arcoleo, K. J., Anson, E. A., et al. (2010). Enduring effects of prenatal and infancy home visiting by nurses on maternal life course and government spending. Archives of Pediatrics & Adolescent Medicine, 164(5), 419 – 424. Olds, D. L., Kitzman, H., Cole, R., Robinson, J. Sidora, K., Luckey, D., et al. (2004). Effects of nurse home visiting on maternal life-course and child development: Agesix follow-up of a randomized control trial. Pediatrics, 114(6), 1550 – 1559. Raphael, S., & Winter-Ebmer, R. (2001). Identifying the effect of unemployment on crime. Journal of Law & Economics, 44(1), 259 – 283. Reynolds, Arthur J. (2000). Success in early intervention: The Chicago Child-Parent Centers. Lincoln: University of Nebraska Press. Reynolds, A. J., Temple, J. A., Robertson, D. L., & Mann, E. A. (2001). Long-term effects of an early childhood intervention on educational achievement and juvenile arrest. JAMA, 285(18), 2339 – 2346. Reynolds, A. J., Temple, J. A., Suh-Ruu, O., Robertson, D. L. Mersky, J. P., Topitzes, J. W., & Niles, M. D. (2007). Effects of a school-based, early childhood intervention on adult health and well-being. Archives of Pediatrics & Adolescent Medicine, 161(8), 730 – 739. Shonkoff, J. P., Garner, A. S., Committee on Psychosocial Aspects of Child and Family Health, American Pediatric Association Committee on Early Childhood, Adoption, and Dependent Care, and Section on Developmental and Behavioral Pediatrics. (2012). The lifelong effects of early childhood adversity and toxic stress. Pediatrics, 129(1), 232 – 246. Spoth, R. L., Redmond, C., & Shin, C. (2000). Reducing adolescents’ aggressive and hostile behaviors: Randomized trial effects of a brief family intervention 4 years past baseline. Archives of Pediatrics & Adolescent Medicine, 154(12), 1248 – 1257. U.S. Department of Housing and Urban Development. (2011, October 14). Moving to Opportunity for Fair Housing demonstration program: Final impacts evaluation. Office of Policy Development and Research. Available at publications/pdf/MTOFHD_fullreport.pdf. U.S. Department of Justice. (2011, September). Arrests in the United States: 1980 to 2009. NCJ 234319. Office of Justice Programs, Bureau of Justice Statistics.

This page intentionally left blank

About the Contributors

Sarah Beth Barnett is Project Officer for the Health Economics and Policy Research and Analysis Team within the Prevention Development and Evaluation Branch in the Division of Violence Prevention at the Centers for Disease Control and Prevention. Eric P. Baumer is Allen E. Liska Professor of Criminology in the College of Criminology and Criminal Justice at Florida State University. Philippe Bourgois is Richard Perry University Professor of Anthropology and Family and Community Medicine at the University of Pennsylvania. Shawn Bushway is Professor of Criminal Justice and Professor of Public Administration and Policy at the University at Albany. His research interests include the economics of crime, desistance, and decision making in the criminal justice system. Philip J. Cook is ITT/Sanford Professor of Public Policy at Duke University. Recent books include Controlling Crime: Strategies and Tradeoffs (coedited) and Paying the Tab: The Economics of Alcohol Policy. Robert D. Crutchfield is Professor and Chair of Sociology at the University of Washington. Linda L. Dahlberg is the associate director for science in the Division of Violence Prevention at the U.S. Centers for Disease Control and Prevention. Dr. Dahlberg has spent much of the past 15 years working in the area of violence prevention — specifically on the efficacy of interventions to reduce violence. Mark Edberg is Associate Professor at the George Washington University School of Public Health and Health Services, with secondary >>





examines race and ethnic differences in rape and sexual assault victimization among U.S. women. Nancy G. Guerra is Professor of Psychology and Associate Dean for Research in the College of Arts and Sciences at the University of Delaware. She also serves as Director of the Global Research Consortium and Editor of the Journal of Research on Adolescence. Karen Heimer is Professor of Sociology and Women’s Studies at the University of Iowa. She conducts research on gender, race, victimization, and offending, as well as on gender and race differences in imprisonment over time. Janet L. Lauritsen is Professor of Criminology and Criminal Justice at the University of Missouri – St. Louis and a visiting research fellow at the Bureau of Justice Statistics, U.S. Department of Justice. Her research is focused on the causes and consequences of victimization and the strengths and limitations of victimization methodologies. Jennifer L. Matjasko is Acting Lead Behavioral Scientist for the Sexual and Youth Violence Evaluation Team within the Prevention Development and Evaluation Branch in the Division of Violence Prevention at the Centers for Disease Control and Prevention. James A. Mercy is Special Advisor in the Division of Violence Prevention at the Centers for Disease Control and Prevention. Matthew Phillips recently received his PhD in criminal justice from the University at Albany. His research interests include drug trafficking, drug-related violence, and quantitative methods. Richard Rosenfeld is Curators Professor of Criminology and Criminal Justice at the University of Missouri – St. Louis. He is past president of the American Society of Criminology, and his current research focuses on the impact of the economy and policing on crime.







observational data, 17 – 18 Order Maintenance Policing (OMP), 235 – 36 ParentCorps, 272 parents: employment, 155, 156, 174; involvement in child’s school, 163t, 167t, 169t, 172 – 73, 173t performative self, 192 – 93 police, 26 police force size, 65t, 70f, 72, 79t policy and incarceration, 238 – 40 policy implications, 293 – 94; strategies that differ by type of economic factors, 294 – 95 policy-relevant data, 18 – 19 positive youth development (PYD) approach, 188 – 89 poverty, 94, 94f; consumer pessimism and, 94, 94f, 96t, 104t; and youth violence, 96t, 103f, 103 – 4, 104t, 106, 106f, 165 poverty measures, 93 poverty traps, two kinds of, 236 – 38 prevention of youth violence: communitylevel, 136; developmental considerations for, 295 – 97; developmental perspective on macroeconomic influences and, 270 – 74; implications for, 293 – 97 prevention strategies within the family context, 318 – 20 problem behavior syndrome theory, 188 property crime: economic activity and, 43t, 45; unemployment and, 54f; youth violence and, 16 – 17 punishment, fear of, 69, 70f, 72 race and ethnicity, 86, 91, 236 – 38, 279 (see also under gun violence); incarceration, family integrity, and, 217 – 19 racial and ethnic categories, 114n5 Raffalovich, Lawrence E., 88 – 89 Raphael, Steven, 28 recessions, defined, 115n12 resilience, 188 risk and protective factor model, 188 – 90, 195 – 96 risk factors: cumulative nature of risk across development, 258 – 70; developmentalstage-specific, 289 – 90, 290t robbery (see also property crime): business cycle and, 32t; economic activity and, 34f, 43t, 45, 47 – 48